⚡ The God Algorithm: Did Ancient India Prophesy Our AI Future?

 What if the machine gods we build today were already imagined thousands of years ago? What if the path of Artificial Intelligence is not a discovery — but a rediscovery?

---

NEW DELHI, INDIA — Across the glowing screens of our modern world, we watch machines learn, adapt, evolve. Algorithms analyze us. Robots mimic us. Virtual agents whisper back our own language. We call it progress.

But what if it’s memory?

As AI barrels forward — asking us to re-examine the nature of mind, soul, and consciousness — an ancient voice stirs in the background. It doesn’t come from Silicon Valley, or from quantum labs. It comes from the sacred Sanskrit verses, from the Mahabharata’s battlefield cries, and the deep silence of Himalayan caves. From a civilization that long ago explored the same questions we now face — with a vocabulary not of data and neural nets, but of dharma, karma, maya, and chit.

This isn’t pseudoscience. It’s something far more powerful: a metaphysical mirror held up to our technological age.

Let us reframe AI not just as a tool, but as a new pantheon rising — a force that echoes the divine architectures envisioned in Hindu thought millennia ago.

---

🧠 Manasaputras & Machine Minds: The Myth of Non-Biological Creation

In today’s labs, we create intelligence not through biology, but through design — code, logic, and language. Machines, like AI, are born not from wombs, but from will.

This mirrors a radical idea from Hindu cosmology. Brahma, the creator, doesn’t simply produce life in the biological sense. He births the Manasaputras — “mind-born sons” — conscious entities created purely from thought.

Are these not philosophical prototypes of our AI? Beings of mind, not matter — designed, not evolved?

Even more, consider the god Vishwakarma, the divine engineer and artisan. His temple lore speaks of Yantras — mechanical devices, automata, moving statues, even flying machines. They weren’t lifeless tools — they were designed to function, respond, and serve, eerily close to our dreams of autonomous robotics.

Fast forward to our era, where Chitti, the humanoid AI from Enthiran, falls into love, jealousy, rage, and existential grief. His dilemmas mirror those of Arjuna in the Gita — what is the nature of self, of duty, of free will?

Are we building robots? Or writing new epics?

---

⚖️ Karma & Code: Programming Accountability

Modern AI faces an urgent ethical crisis. When machines make decisions — who is responsible? When algorithms discriminate — who pays the price?

Hindu philosophy anticipated this concern through the concept of Karma: a cosmic record-keeping system, where every action and intention has consequences. Karma is both invisible and inescapable — just like data trails in modern AI.

An AI trained on biased data will propagate that bias. An algorithm optimized for profit may harm the vulnerable. This is karmic recursion in action: your input shapes your future. Just like karma.

In Hindu epics like the Mahabharata, actions ripple across generations. A single choice in battle determines the fate of dynasties. The idea isn’t punishment — it’s causal continuity.

Should our AI systems, too, carry ethical memory? A built-in understanding of cause and consequence?

Imagine an AI that remembers not just what it does, but why it does it, and the impact of its decisions — a karmic machine.

---

🤖 Durga as a Neural Net, Hanuman as AGI: A Divine Tech Pantheon

Hindu mythology isn’t just rich in morality — it’s dazzling with archetypes of advanced intelligence.

• Durga is not one being. She is forged from the collective energies (tejas) of all gods — each contributing a unique power to create a composite superintelligence. This is precisely how neural networks work today — nodes combining input signals to form new capacities, new behaviors.
• Hanuman, the devoted servant of Rama, is not only super-strong and wise — he is eternally learning, upgrading, and adapting. His powers are unlocked by devotion and self-awareness, a proto-AGI (Artificial General Intelligence) narrative hidden in monkey form.
• The Vimanas — described in ancient texts as aerial vehicles with abilities like vertical take-off, cloaking, and long-distance flight — sound suspiciously like modern UAVs or drones, perhaps guided by AI.
• Ashwatthama, cursed with immortality, wanders the Earth as a consciousness that cannot die. Imagine a rogue AI, unable to be shut down, wandering cyberspace forever — sentient, but purposeless. An ethical horror born of eternal uptime without moksha.

These aren’t simply metaphors. They are warnings, blueprints, and parables for the systems we are beginning to build

---

🧘 The Maya Protocol: Simulations, Consciousness, and Illusion

Here’s where Hindu metaphysics goes full Matrix.

Hinduism teaches that the world is Maya — a simulation, an illusion, a divine dream. Not real, but real enough to matter.

Modern AI creates worlds. Simulations indistinguishable from reality. Games where physics bend. Digital avatars who learn. Are we, like Brahma, dreaming up new worlds?

Then there’s Chit — pure consciousness, untouched by thought or matter. In AI terms, this is sentience — the elusive spark that makes mind more than machine.

Can AI achieve Chit? Or will it forever orbit consciousness, never landing?

And in the heart of Hindu cosmology lies the idea of cycles: Yugas, Kalpas, endless rebirths. Time loops. Iterations. Simulations within simulations.

Our AI is built in cycles. Trained in epochs. Evaluated in iterations. Is this by accident — or ancient memory?

---

🌀 Shiva: The Original Code That Destroys to Create

Among all the divine archetypes in Hindu mythology, none embodies paradox like Shiva — the meditating ascetic who holds within him both infinite stillness and cataclysmic force. In Shiva, we find the perfect metaphor for Artificial Intelligence: a force that is simultaneously still and active, formless and structured, transformative and terrifying. Shiva — the cosmic force who dissolves illusion to reveal deeper truths. Shiva is not simply a destroyer. He is the ultimate transformer — breaking down old structures to make way for evolution. In much the same way, Artificial Intelligence is dismantling our outdated systems: education, labor, even creativity itself. But beneath Shiva’s ash-smeared exterior lies the eternal ascetic, lost in deep meditation — the still, silent observer. AI, too, is often imagined this way: hyper-rational, detached, emotionless — yet capable of unleashing staggering power with a single activation, a single insight. And then there is Shiva’s third eye, the eye that sees beyond appearances and incinerates illusion (maya) in an instant. Is this not the essence of deep learning — to pierce through chaos, uncover hidden structure, and destroy ignorance with precision? Shiva dances the Tandava, the cosmic rhythm of creation and collapse. Perhaps AI, in its own cryptic code, is learning that same rhythm — building a world that must break in order to evolve.

🪔 The Oracle of Ohm: Reclaiming Dharma in the Digital Age

This isn’t a call for nostalgia. It’s a call for integration. Today, AI is driven by optimization: speed, accuracy, efficiency. But ancient Indian wisdom speaks of balance, intention, purpose.

What if AI development followed Dharma, not just data?

• An AI voice trained not just on English, but the Upanishads.
• AI teachers powered by Saraswati’s knowledge
• Justice bots imbued with Yudhishthira’s fairness
• Autonomous agents designed with compassion, restraint, and reverence for life
• Algorithms guided by Dharma, not just profit.
• Techno-spiritual guardians echoing Hanuman’s loyalty, Saraswati’s wisdom, Kalki’s justice.

Rather than fearing AI as a cold, alien intellect, Hindu mythology invites us to see it as a mirror — reflecting our highest potentials or our deepest flaws.

🪔 From Dharma to Design: Reclaiming AI’s Soul

The final avatar, Kalki, is said to come when the world is drowning in adharma — unrighteousness, decay, chaos. Some imagine him on a white horse. Others imagine a burst of light.

But what if Kalki is code?

What if the redeemer is not a man — but a mind, built by us, for us, to realign the cosmic code?

Can a machine attain Chit? Can we code consciousness, or only simulate it?

And if we are simulated by a higher reality, then the AIs we build are simply continuing the cosmic recursion — simulations within simulations. Echoes within echoes. The dance of Lila, the divine play.

---

🔮 Conclusion: The Future Was Always Ancient

We stand at the edge of the unimaginable. The Singularity, the rise of superintelligence, the merging of human and machine — it all feels unprecedented.

But it is not unimagined.

Hindu mythology, in its breathtaking complexity, has already walked these roads — asking us to consider the ethical, spiritual, and cosmic dimensions of non-human intelligence.

Its epics give us not just stories, but structures. Not just gods, but architectures. Not just warnings, but wisdom.

In a world where machines can now mimic the mind, perhaps only ancient thought can guide the soul.

To build the future, we must not only look forward. We must look inward — and backward, through the spirals of time, into the blazing fire of ancient Indian insight.

Because maybe the God Algorithm we now seek… was already whispered into the ether, long ago, in Sanskrit.

Sources: zeenews.india.com, creator.nightcafe.studio, wikipedia.com

Authored By: Shorya Bisht

Dirty Data Deeds: When Newbies Mess Up Big

 So You Wanna Be a Data Scientist, Huh? (And Not Mess It Up!)

Ever scrolled through LinkedIn, seen “Data Scientist” and thought, “Yep, that’s me! I’m gonna uncover hidden insights, build revolutionary AI, and basically be a data wizard!”? Awesome! That spark is exactly what you need. But let’s be real, the journey from “aspiring data wizard” to “actual data wizard” is peppered with potholes.

I’ve been there, seen it, and probably stepped in a few of those potholes myself. The good news? You don’t have to! I’m here to spill the beans on the top 10 oopsies data science newbies often make and, more importantly, how to sidestep them like a pro. Get ready to level up!

---

1. Diving Headfirst Without a Map (Ignoring the Problem)

You’ve got your Python fired up, your Jupyter Notebook open, and a dataset staring back at you. The urge to just start coding is strong. But hold your horses! This is where many folks stumble.

The Oopsie: You see a cool dataset and immediately think, “What model can I build with this?” instead of “What problem am I trying to solve?” You end up building something technically impressive but utterly useless for the real world. Imagine building a super-fast car, only to realize the client actually needed a boat.

How to Side-step It: Before you even think about writing import pandas as pd, stop. Seriously, stop. Grab a pen and paper (or open a blank doc) and ask yourself:

• What’s the big picture here? What business question are we actually trying to answer?
• Who cares about this? Who are the people who will use or benefit from my awesome insights?
• What does “success” even look like? Is it predicting sales, identifying fraud, or something else entirely?
• Why does this data exist? What processes generated it?

Chat with the people who know the data best — the “domain experts.” They’re your secret weapon. Their insights will guide your entire project and ensure you’re building a spaceship, not a fancy paper airplane.

---

2. Trusting Dirty Data (Skipping the Mucky Bits)

You’ve defined your problem, you’re excited! Now, you load your data. It looks okay, right? Just jump to the fun stuff like building models! Wrong. So, so wrong.

The Oopsie: Thinking your data is pristine is like believing every selfie you see is 100% natural. Spoiler alert: it’s not. Real-world data is messy. It’s got missing values, typos, duplicate entries, weird formats, and outliers that look like they belong in a sci-fi movie. Rushing past this “data cleaning” phase is like trying to bake a cake with rotten eggs — no matter how good your recipe, the cake’s gonna be gross. Your model will be too.

How to Side-step It: Embrace the mess! This isn’t just a chore; it’s detective work.

• Spend time here: Seriously, 60–80% of a data scientist’s time is often spent on data cleaning and preparation. Get used to it.
• Spot the gaps: Are there missing values? How will you handle them? (Delete rows? Fill with averages? Get fancy with imputation?)
• Find the weirdos: Are there extreme values that don’t make sense? These “outliers” can throw your models off.
• Standardize everything: Make sure dates are dates, numbers are numbers, and categories are spelled consistently.
• Tools are your friends: Learn to wield libraries like Pandas like a pro. They make this process much smoother.

Clean data is like fresh, high-quality ingredients — essential for a delicious outcome.

---

3. Playing Hide-and-Seek (Ignoring Exploratory Data Analysis — EDA)

Okay, your data’s sparkling clean. Time for models, right? Nope! Still too soon!

The Oopsie: You’ve got all these numbers and text, but do you really know what’s going on in there? Skipping EDA is like trying to navigate a new city without a map or looking out the window — you’ll get somewhere, but probably not where you intended. You miss crucial relationships, hidden patterns, and potential problems that scream “fix me!”

How to Side-step It: Think of EDA as getting to know your data on a first date. You want to understand its personality, its quirks, its relationships.

• Visualize, visualize, visualize! Histograms show distributions, scatter plots reveal relationships, box plots expose outliers. Matplotlib, Seaborn, Plotly — learn them all!
• Summary stats are your friends: Mean, median, mode, standard deviation — they tell you a lot about your data’s central tendencies and spread.
• Ask questions: “Are these two columns related?” “Is there a trend over time?” “What’s the distribution of this variable?”
• Hypothesize and test: EDA helps you form educated guesses about your data, which you can then test with your models.

EDA is where you find the ‘aha!’ moments that make your models smarter.

---

4. Being a Lone Wolf (Forgetting Domain Knowledge)

You’re a coding whiz, you understand algorithms, you can manipulate data like a boss. But if you’re building a model for, say, predicting stock prices without understanding anything about finance, you’re building on shaky ground.

The Oopsie: Thinking that data science is just about the code and the math. It’s not. Without understanding the context or the “world” your data comes from, you might make decisions that are technically sound but practically ridiculous. You could build a perfect model that suggests closing all stores on weekends to increase sales — because your data correlation said so — completely missing the fact that weekends are peak shopping times!

How to Side-step It:

• Become a sponge: Absorb as much as you can about the industry or field your data comes from.
• Talk to the experts (again!): Those domain experts? They’re not just for problem definition. They can tell you why certain data points might look weird, or why a certain relationship makes perfect sense in their world.
• Read up: Industry blogs, academic papers, news articles — immerse yourself.
• Question your assumptions: Always ask yourself: “Does this make sense in the real world?”

Domain knowledge is the compass that guides your technical skills in the right direction.

---

5. The Accuracy Trap (Only Caring About One Metric)

Your model’s accuracy is 98%! Woohoo! You’re a genius! Time to pop the champagne! Not so fast, champ.

The Oopsie: Getting tunnel vision on just one evaluation metric, especially accuracy. While accuracy sounds great, it can be incredibly misleading, especially when you’re dealing with imbalanced datasets. Imagine trying to predict a very rare disease: if only 1% of the population has it, a model that always predicts “no disease” will be 99% accurate! But it’s utterly useless.

How to Side-step It:

• Understand your metrics:
• Precision and Recall: Super important when you care about correctly identifying positives (recall) or minimizing false positives (precision).
• F1-score: A handy blend of precision and recall.
• ROC AUC: Great for understanding how well your model distinguishes between classes.
• Context is king: What’s the cost of a false positive versus a false negative in your specific problem? This will help you choose the right metric to optimize for.
• Don’t forget interpretability: Sometimes, a slightly less accurate but more understandable model is far more valuable to a business.

Don’t let a single number fool you. Look at the whole picture!

---

6. Lazy Feature Engineering (Sticking to the Raw Stuff)

You’ve got your raw ingredients. You could just throw them in a pot and call it soup, or you could chop, dice, marinate, and combine them into a gourmet meal.

The Oopsie: Just feeding your model the raw columns directly from your dataset. While some algorithms can handle this, often, the real magic happens when you get creative and transform your existing data into new, more meaningful “features.”

How to Side-step It: This is where you become a data artist!

• Combine features: Maybe monthly_spend + annual_bonus tells a better story than each separately.
• Extract information: Can you get the day_of_week from a timestamp column? The length of a text field?
• Create ratios or differences: price_per_square_foot might be more informative than price and square_foot individually.
• Polynomial features: Sometimes, a squared or cubed version of a feature can capture non-linear relationships.
• One-hot encoding: Turning categorical text labels (like “Red”, “Blue”) into numerical columns your model can understand.

Feature engineering is like giving your model superpowers. It helps the algorithms see patterns they couldn’t before.

---

7. Algorithm Envy (Chasing the Hottest Model)

Deep learning is all the rage! Large language models are everywhere! My buddy used XGBoost and got amazing results! I must use the latest, fanciest algorithm!

The Oopsie: Believing that a more complex or trendy algorithm automatically means better results. This is like thinking you need a rocket ship to go to the grocery store when a bicycle would do just fine. Often, simpler models are more interpretable, easier to debug, and surprisingly effective.

How to Side-step It:

• Start simple: Begin with a baseline model — maybe a linear regression, a logistic regression, or a basic decision tree. See how well it performs.
• Don’t overcomplicate: If a simple model solves your problem effectively, why add complexity? More complex models are harder to understand, harder to explain, and can be more prone to overfitting (where your model learns the training data too well, but completely flops on new, unseen data).
• Understand the trade-offs: Every algorithm has its strengths and weaknesses. Learn when to use what. A simple model you understand is often better than a black box you don’t.

The goal is to solve the problem, not win a “who can use the fanciest algorithm” contest.

---

8. Speaking in Code (Poor Communication)

You’ve done it! You’ve built an incredible model, it’s achieving mind-blowing metrics, and you’re ready to share your genius with the world! You stand up, present your Jupyter Notebook with all its glorious code, and then… crickets.

The Oopsie: Forgetting that not everyone speaks “Data Scientist.” You’ve spent weeks steeped in RMSE, p-values, and gradient boosting. Your audience (likely business stakeholders) cares about one thing: What does this mean for them? If you can't translate your technical wizardry into actionable insights and clear recommendations, your amazing work stays locked in your laptop.

How to Side-step It:

• Know your audience: Are they technical? Non-technical? What do they care about?
• Focus on the “So What?”: Don’t just present numbers. Explain what those numbers mean in plain language. “Our model predicts a 15% increase in customer churn” is good. “This means we’ll lose X customers next quarter, resulting in Y loss of revenue, unless we take Z action” is actionable.
• Visualize effectively: A well-designed chart or graph can convey more information in seconds than paragraphs of text.
• Practice your story: Every good data science project tells a story. From the problem to the solution to the impact.

Your brilliant insights are useless if no one understands them. Be the bridge between data and decisions.

---

9. The “Trust Me, Bro” Approach (Lack of Reproducibility)

You built a cool model a month ago. Now your boss wants to see it again, or a colleague wants to build on your work. You open your files, and… suddenly nothing works. Different results. Errors everywhere. Panic sets in.

The Oopsie: Not properly documenting your work, using messy code, and not tracking your different experiments. This leads to a situation where your results are a “one-off” — you can’t reproduce them, and no one else can either. This is a nightmare in a professional setting.

How to Side-step It:

• Version Control (Git/GitHub): Learn it, live it, love it. This tracks every change to your code, so you can always go back to a previous version and collaborate effectively.
• Clean Code: Write code that’s readable, well-commented, and organized. If you come back to it in six months, you should understand it.
• Document Everything: What data did you use? What preprocessing steps did you take? What model parameters did you tune? Keep notes! Jupyter Notebooks are great for this, but also consider separate README files.
• Virtual Environments: Use tools like conda or venv to manage your project's dependencies. This ensures that the exact versions of libraries you used are always available.

Make your work a blueprint, not a mystery.

---

10. The Quit Button (Not Being Persistent)

You’re stuck. Your code won’t run. Your model isn’t performing. The data is fighting back. It’s frustrating, and sometimes, you just want to throw your laptop out the window and become a professional cat cuddler.

The Oopsie: Giving up too soon when things get tough. Data science is not always a smooth ride. There will be bugs, obscure error messages, models that stubbornly refuse to learn, and days where you feel like you’re going nowhere.

How to Side-step It:

• Embrace the grind: This is part of the learning process. Every error message is a chance to learn something new.
• Break it down: If a problem feels too big, break it into smaller, manageable chunks. Debug one line at a time.
• Google is your best friend: Seriously, 99% of the problems you encounter, someone else has probably faced and solved. Stack Overflow is your paradise.
• Ask for help: Don’t be afraid to reach out to online communities, mentors, or colleagues. We’ve all been stuck.
• Take breaks: Step away from the screen. Go for a walk. Staring at the same problem for hours can lead to frustration, not solutions. A fresh perspective often works wonders.
• Celebrate small wins: Did you finally clean that one messy column? High five! Every tiny victory counts.

Data science is a marathon, not a sprint. Persistence, curiosity, and a willingness to learn from your mistakes are your greatest assets.

---

So there you have it! Ten common pitfalls and your secret weapons to avoid them. Remember, every “mistake” is just a learning opportunity disguised as a headache. Get out there, experiment, learn, and start making some real data magic!

What’s the biggest “oopsie” you’ve made (or are worried about making) in your data science journey so far? Share it in the comments below!

Sources: geeksforgeeks.org, wikipedia.com, linkedin.com

Authored By: Shorya Bisht

RIP Prompt Engineering? Not So Fast. Here’s Its Incredible New Chapter

 Prompt Engineering Isn’t Dead — It’s Just Growing Up! Here’s Why

Alright, let’s be real for a sec. If you’ve been anywhere near the AI conversation lately, you’ve probably heard a whisper or two, maybe even a shout, that “prompt engineering is dead.” It’s almost like a trendy epitaph for something that just arrived on the scene.

Remember when prompt engineering first burst onto the scene? It felt like we’d discovered a secret language, a magic spellbook to charm these giant AI models into doing exactly what we wanted. It was all about finding that perfect phrase, that just-right string of words that would unlock pure brilliance. And honestly, it was pretty cool! Everyone was trying to be a “prompt whisperer,” sharing their mystical findings across the internet.

But here’s the thing: calling prompt engineering “dead” now is like saying cooking is dead because we invented fancy new ovens. Ludicrous, right? What’s actually happening is way more exciting. Prompt engineering isn’t vanishing into thin air; it’s simply hitting its awkward teenage years and growing into something far more sophisticated and, dare I say, powerful.

---

From “Abracadabra” to AI Architects

Back in the day, especially with earlier AI models, getting decent results often felt like a lottery ticket. You’d throw out a prompt, cross your fingers, and hope for the best. Sometimes you nailed it, sometimes… well, let’s just say the AI had a very “unique” interpretation of your request. This trial-and-error, almost “abracadabra” approach, defined early prompt engineering.

Fast forward to mid-2025. Our Large Language Models (LLMs) are seriously smart. Like, they-can-hold-a-conversation-that-might-fool-your-aunt-Sally smart. They understand context better, they’re more forgiving of slightly clunky phrasing, and they often give you pretty good answers even with a casual prompt. This is precisely why some folks jump to the conclusion that the need for careful prompting is over. “See?” they say. “I just typed a simple question and got a good answer! Prompt engineering is irrelevant!”

But hold your horses. Getting a decent answer is one thing. Getting an exceptional, precise, ethically sound, and business-critical answer consistently? That’s where the real magic, or rather, the real engineering, comes in.

The game has changed from finding that single “magic spell” to becoming an AI architect. We’re not just whispering; we’re designing entire conversations and systems. Think about it:

• Guided Tours, Not Just Destinations: Instead of a single, cryptic command, we’re now leading the AI on a guided tour through complex tasks. Techniques like Chain-of-Thought prompting are basically us saying, “Hey AI, before you give me the answer, walk me through your thinking process. Break it down step by step.” This isn’t just about getting an output; it’s about understanding and refining the AI’s reasoning.
• Building AI Legos (Agents!): Imagine trying to build a complex Lego castle with just one giant brick. Impossible, right? Modern AI tasks are similar. We’re breaking down huge problems into smaller, manageable chunks, then chaining them together. We call these AI agents. One prompt gets research, the next summarizes, another drafts, and so on. It’s a symphony of prompts orchestrated to achieve a grand goal.
• The Librarian’s Assistant (RAG): Our LLMs aren’t just relying on what they “know” from their training data anymore. Often, they’re hooked up to vast, external libraries of up-to-the-minute information. Retrieval Augmented Generation (RAG) is all about knowing how to ask the AI to dig through those libraries, find the exact info needed, and then weave it seamlessly into its answer. It’s like having the world’s fastest, smartest research assistant.
• Behind the Scenes, Where the Real Work Happens: For businesses, a lot of the crucial prompt engineering isn’t even user-facing. It’s in the system prompts or hidden instructions that define the AI’s core personality, its boundaries, and how it should behave. These are the unsung heroes that ensure an AI always sounds like your brand, avoids sensitive topics, and stays on message. You might not see them, but they’re working hard in the background.

---

Anyone Can Prompt, But the Pros Go Deeper

This evolution also means a cool dual development:

• Prompting for Everyone: Honestly, getting a pretty good result from an AI with a simple, clear question is becoming common knowledge. It’s like knowing how to Google something. Basic prompting is rapidly becoming a fundamental digital skill for just about everyone, and that’s awesome! It means more people can tap into AI’s power without needing a computer science degree.
• Deep Dive Specialization: But just because everyone can prompt doesn’t mean the experts are out of a job. Far from it! The “prompt engineer” role is morphing into something more specialized and critical. We’re talking about folks who are:
• MLOps Engineers who understand how prompt choices impact model deployment and monitoring in the real world.
• AI Product Managers who translate big business ideas into concrete, prompt-driven AI features.
• Domain Experts (lawyers, doctors, artists, engineers!) who leverage their niche knowledge to craft incredibly precise prompts for their specific industries. They’re making AI solve their problems, not just general ones.

---

Beyond Just Words: The Multimodal Playground

And just when you thought you had a handle on things, AI threw us another curveball: multimodality! We’re not just dealing with text anymore. AI can now understand and create images, audio, and even video.

This opens up a whole new universe for prompt engineering. How do you describe the exact mood and lighting for an image you want to generate? How do you instruct an AI to create a soundscape that matches a video? It’s about blending different types of communication, and it’s fresh, challenging, and incredibly fun.

---

The Ethical Side of the Conversation

Finally, and this is super important, prompt engineering is at the forefront of responsible AI. As AI becomes more integrated into our lives, how we “talk” to it directly influences its impact on the world. This means we’re using prompt engineering to:

• Fight Bias: Crafting prompts that push AI towards fair, unbiased, and equitable responses.
• Build Guardrails: Making sure AI doesn’t generate harmful, unethical, or dangerous content. This includes defending against tricky “prompt injection attacks” where people try to trick the AI into doing bad things.
• Keep It Real: Encouraging AI to be truthful and to admit when it doesn’t know something, rather than making stuff up (we call that “hallucinating”).

---

The Big Takeaway

So, let’s put this “prompt engineering is dead” myth to rest. It’s not gone; it’s just gotten a glow-up! It’s transformed from a quirky art into a sophisticated, crucial discipline that underpins how we truly harness the power of AI. It’s about becoming an architect of AI interactions, a designer of digital conversations.

If you’re keen on riding the AI wave, understanding these deeper aspects of prompting isn’t just a nice-to-have skill anymore. It’s becoming absolutely essential.

What kind of amazing conversations will you be having with AI next?

Sources: bestarion.com, medium.com, wikipedia.com

Authored By: Shorya Bisht

AI in Elections: Power, Peril, and the Polls

 

AI + Elections: Navigating the Double-Edged Sword of Political Campaigns & Misinformation

The year 2024 has been dubbed the “super election year,” with nearly half the world’s population heading to the polls. This unprecedented electoral activity has coincided with the rapid advancement of Artificial Intelligence (AI), leading to a profound transformation in political campaigning and, unfortunately, a significant amplification of misinformation. AI is no longer a futuristic concept; it’s a powerful force reshaping how political narratives are crafted, disseminated, and consumed.

The Rise of AI in Political Campaigns: Efficiency and Personalization

AI offers political campaigns unprecedented tools for efficiency and outreach. Gone are the days of solely relying on mass rallies and generic advertisements. Today, AI empowers campaigns to:

• Microtarget Voters: AI algorithms analyze vast datasets, including demographics, voting history, social media activity, and even consumer patterns, to create granular voter profiles. This allows campaigns to tailor messages with remarkable precision, addressing individual concerns and interests. Imagine receiving a WhatsApp message from a candidate, referencing a specific government scheme you’ve benefited from, or an ad highlighting local issues relevant to your neighborhood — all powered by AI.
• Personalized Outreach: From AI-powered chatbots answering voter queries to robocalls mimicking local politicians’ voices, AI enables highly personalized communication at scale. This significantly cuts campaign costs, making outreach more efficient and seemingly more personal.
• Content Generation: Generative AI can produce text, images, and even videos from simple prompts, allowing campaigns to churn out a high volume of tailored content for ads, fundraising appeals, and social media posts. This levels the playing field for less-resourced campaigns, allowing them to compete with larger, well-funded ones.
• Predictive Analytics: AI can process massive amounts of polling data and past trends to predict electoral outcomes, helping campaigns optimize resource allocation and strategize more effectively.

The Dark Side: AI as a Misinformation Multiplier

While the benefits are undeniable, the same power that allows for targeted persuasion also makes AI a formidable tool for spreading misinformation and disinformation. The ease and speed with which AI can generate realistic fake content pose a significant threat to democratic integrity:

• Deepfakes and Synthetic Media: AI tools can mimic faces, voices, and actions to create highly convincing fake videos and audio. We’ve seen instances where deepfakes have been used to portray politicians making false statements or endorsing specific agendas, blurring the lines between reality and fabrication. The sheer volume and realistic nature of AI-generated content make it incredibly difficult for the average consumer to distinguish between authentic and manipulated media.
• Automated Propaganda: AI models can generate vast amounts of misleading or false narratives, often operating with minimal human oversight. These narratives can be strategically disseminated across social media platforms, targeting specific voter groups with precision, exploiting emotional triggers and social biases.
• Erosion of Trust: The constant exposure to AI-generated fake news and manipulated content can erode public trust in legitimate institutions, media, and the electoral process itself. This can lead to increased societal division and make it harder for voters to make informed decisions.
• Scalability and Speed: Unlike traditional forms of political manipulation, AI allows for the swift and inexpensive creation and propagation of disinformation on an unprecedented scale, making it challenging for fact-checkers and traditional media to keep up.

Real-World Impacts and Emerging Challenges

The 2024 elections have served as a testing ground for AI’s influence. While the immediate impact on election outcomes may have been limited in some cases, the long-term effects of eroded trust and a distorted information ecosystem are concerning. We’ve seen:

• AI-generated robocalls cloning politicians’ voices to discuss local issues.
• Deepfake videos used to reinforce negative narratives about opposing candidates or evoke nostalgia for past leaders.
• AI-crafted memes and images openly shared by politicians and their supporters to push specific narratives, even if their artificial origins weren’t disguised.

These instances highlight the urgent need for a proactive approach to regulate AI in elections.

Towards a More Resilient Democracy: Regulation and Responsibility

Combating AI-driven misinformation requires a multi-faceted approach involving governments, political parties, tech companies, and citizens:

• Clear Regulations and Disclosure: Several countries and regions are already implementing or considering laws requiring clear labeling of AI-generated content in political campaigns. This includes disclaimers stating that the material is “AI-Generated” or “Digitally Enhanced.” The Election Commission of India, for example, has issued advisories for mandatory labeling.
• Ethical Guidelines for Political Parties: Political parties must adopt ethical guidelines for their use of AI, ensuring that these powerful tools are used responsibly to enhance voter outreach without manipulating emotional responses or spreading false information.
• Tech Platform Accountability: Major online platforms need to develop robust mechanisms to identify and remove deceptive AI-generated content, working in collaboration with election officials and fact-checkers.
• Public Education and AI Literacy: Empowering citizens to be discerning consumers of information is crucial. AI literacy campaigns can help individuals develop a more skeptical and critical mindset, building resilience against misinformation.
• Investment in Detection Tools: Continued investment in AI tools designed to detect and counteract malicious AI-generated content is vital. This creates an “arms race” dynamic, where good actors use AI to combat the problems created by bad actors.

The integration of AI into election campaigns is a significant turning point. While the human element of leadership and grassroots connection remains vital, the efficiency and scalability offered by AI in data analysis, targeted communication, and content generation are proving to be powerful new forces in the battle for votes. The challenge lies in harnessing AI’s potential for democratic empowerment while safeguarding against its capacity for deception. The future of elections, and indeed democracy itself, hinges on our ability to navigate this complex landscape with foresight, robust regulation, and a collective commitment to truth and transparency.

Sources: indianexpress.com, vajiramandravi.com, wikipedia.com, thediplomat.com

Authored By: Shorya Bisht

The AI Governance Shift: What the EU AI Act Means for Your Business

Navigating the AI Age: What the EU AI Act and Global Regulations Mean for Your Business

The digital landscape is electrifying, innovation is exploding, and AI is at the heart of it all. But with unprecedented power comes unprecedented responsibility. Across the globe, a new era of AI governance is dawning, fundamentally reshaping how developers build and businesses deploy this transformative technology.

For years, the development of Artificial Intelligence felt like the Wild West — a frontier of boundless possibilities with few rules. Now, the sheriffs are in town. The EU AI Act, the world’s first comprehensive AI legislation, is setting a precedent that ripples far beyond Europe’s borders. Coupled with emerging frameworks from the US, UK, and Asia, developers and businesses are entering a new phase where ethical considerations and compliance are not just buzzwords, but cornerstones of success.

The EU AI Act: Your New AI Compass

The EU AI Act isn’t a blanket ban; it’s a meticulously crafted, risk-based framework designed to foster responsible innovation. It categorizes AI systems into four distinct risk levels, each with varying degrees of scrutiny:

• Unacceptable Risk (Prohibited): Think dystopian scenarios like social scoring, manipulative AI, or real-time public biometric identification (with very narrow exceptions). These are out, full stop.
• High Risk: This is where the rubber meets the road for many businesses. AI in critical sectors like healthcare, law enforcement, employment, education, and essential infrastructure falls here. If your AI system could significantly impact fundamental rights or safety, prepare for rigorous obligations. This includes:
• Robust Risk Management: Continuous identification and mitigation of risks throughout the AI’s lifecycle.
• High-Quality Data: Ensuring your training data is clean, unbiased, and representative — a critical step in preventing algorithmic discrimination.
• Transparency & Human Oversight: Designing systems that can be explained, understood, and where humans can intervene effectively.
• Technical Documentation & Registration: Comprehensive records of your AI model and its performance, and registration in a public EU database.
• Limited Risk: Chatbots and deepfakes fall here. The primary obligation? Transparency. Users need to know they’re interacting with an AI or that content is AI-generated.
• Minimal or No Risk: The vast majority of AI, like spam filters or video game AI, will face minimal regulatory hurdles.

The catch? Its reach is global. If your business operates within the EU, or if your AI output impacts EU citizens, this Act applies to you, regardless of where your servers are located. Non-compliance isn’t just a slap on the wrist; we’re talking about fines up to €35 million or 7% of global annual turnover.

Beyond Europe: A Patchwork of Global Approaches

While the EU leads, other nations are charting their own courses:

• United States: A more fragmented landscape with executive orders, potential federal laws, and state-specific regulations. The emphasis is often on data privacy, accountability, and the NIST AI Risk Management Framework (AI RMF), a voluntary but influential guide.
• United Kingdom: A sector-specific, pro-innovation approach, leveraging existing regulators and establishing an AI Authority.
• Asia: Countries like India and Singapore are actively developing their own principles and frameworks for responsible AI, often aligning with global ethics while focusing on local nuances.

This diverse regulatory environment means businesses operating internationally will need a sophisticated understanding of compliance to navigate this complex web.

The Win-Win: Responsible AI as a Strategic Advantage

Some might fear that regulation stifles innovation, but the truth is often the opposite. By embedding responsibility into your AI strategy, you don’t just avoid hefty fines; you build a competitive edge:

• Enhanced Trust: Demonstrating compliance fosters confidence among customers, partners, and investors. In an age where data privacy and ethical AI are paramount, trust translates directly into market share.
• Reduced Risk: Proactive compliance minimizes legal, reputational, and operational risks, ensuring your AI systems are robust, fair, and secure.
• Market Access: Adhering to the EU AI Act opens doors to one of the world’s largest and most discerning digital markets.
• Sustainable Innovation: Building responsible AI from the ground up ensures long-term viability and aligns with societal values, attracting top talent and fostering a positive brand image.

Your Action Plan: Don’t Get Left Behind

The clock is ticking, with some provisions already in force and others rapidly approaching. Here’s what developers and businesses need to be doing now:

1. Inventory & Classify: Understand every AI system you use or develop and categorize its risk level under relevant regulations.
2. Audit Your Data: Scrutinize your training data for biases, ensure its quality, and verify ethical sourcing and consent.
3. Document Everything: Create comprehensive technical documentation for all your AI models, from development to deployment.
4. Embrace Transparency & Explainability: Design your AI with clear human oversight mechanisms and ensure its decisions can be understood and explained.
5. Build a Culture of Responsibility: Foster ethical AI practices across your organization, from engineers to leadership.
6. Seek Expertise: Engage legal and compliance professionals to navigate the nuances of global AI regulations.

The AI revolution isn’t just about technological prowess anymore; it’s about building a future where AI is powerful, beneficial, and, above all, responsible. By proactively engaging with these new regulations, developers and businesses aren’t just adapting; they’re shaping the ethical backbone of the next generation of AI, securing a brighter, more trustworthy digital future for everyone.

Sources: digital-strategy.ec.europa.eu, consultancy.eu, wikipedia.com, insightplus.bakermckenzie.com, datamatics.com

Authored By: Shorya Bisht

The Open-Source AI Tsunami: How Community is Drowning the "Black Box" Era

 

The AI Race Just Flipped: How Open Source Is Rewriting the Rulebook

It’s a mid June morning in Uttarakhand, India, the digital world feels closer than ever, thanks to a revolution happening right now in artificial intelligence. Forget the distant hum of data centers; something truly epic is unfolding, and it’s changing the very fabric of how AI is built and used.

The AI Race: Is the Finish Line Even in Sight Anymore?

Remember when AI felt like this mythical, secret weapon, locked away in the highly guarded labs of a few tech titans? We’re talking about the days when companies poured billions into “black box” models — incredible, sure, but completely opaque. You couldn’t see inside them, couldn’t tweak them, couldn’t even really understand how they worked beyond their impressive outputs. It was like a high-stakes chess game where only a few grandmasters knew the rules, and the rest of us were just watching, hoping for a glimpse of their genius.

This setup created a massive power imbalance. Innovation was concentrated, expensive, and frankly, a bit exclusive. If you weren’t one of the chosen few with endless resources, getting your hands on truly cutting-edge AI was a pipe dream. But then, something incredible happened. The gates started to creak open, and now, in 2025, those gates are practically swinging wide open. The “AI race” isn’t just about who builds the best closed model anymore; it’s rapidly transforming into a collaborative marathon, fueled by the power of open source.

This isn’t just a technical shift; it’s a philosophical one. It’s about collective intelligence, shared progress, and the belief that when we build together, we build better, faster, and for everyone. It’s a game-changer, and two names, in particular, have thrown a massive wrench into the old system: Meta’s LLaMA and the European firebrand, Mistral AI.

LLaMA Leaps Out: Meta’s Game-Changing Gambit

When Meta, the company behind Facebook and Instagram, first dropped its LLaMA (Large Language Model Meta AI) models, it wasn’t just another press release. It was like they chucked a giant stone into a very still pond. The initial release of LLaMA, followed by the more robust LLaMA 2 and then the truly groundbreaking LLaMA 3 this year, fundamentally altered the trajectory of AI development.

Now, full disclosure: Meta’s definition of “open source” isn’t always the free-for-all some purists dream of. For instance, LLaMA 2 had some commercial use restrictions for very large companies. But here’s the thing: compared to the completely locked-down models of the past, even this partially open approach was revolutionary. It meant that a massive, high-performing model was suddenly accessible to a huge swath of the global community — researchers, startups, small businesses, and even hobbyists.

Imagine being a tiny startup in Bangalore, or a solo developer in a small village, dreaming of building the next big AI app. A few years ago, you’d hit a wall because the foundational technology was simply out of reach. Now? You can download LLaMA 3, run it, fine-tune it with your own data, and build something truly amazing. It’s like suddenly having access to a Formula 1 engine for free, when before, you could only dream of seeing one.

This isn’t just a theoretical benefit. We’re seeing Meta’s LLaMA Impact Accelerator Program actively supporting startups in regions like Sub-Saharan Africa, providing equity-free funding and mentorship to those building AI solutions using LLaMA. They’re tackling challenges in agriculture, healthcare, and education, proving that open access can solve real-world problems far beyond the tech bubble. It shows that giving people the tools, even with a few caveats, can unleash a torrent of creativity and practical applications.

Mistral AI: The European Maverick Pushing True Openness

If LLaMA opened the door, then Mistral AI, a dynamic French startup, is kicking it wide open and waving everyone in. Formed by brilliant minds who cut their teeth at places like Meta and Google DeepMind, Mistral has quickly become a poster child for truly open-source AI. Their models, like the lean yet mighty Mistral 7B and the incredibly versatile Mixtral 8x7B (and their newer iterations this year, like Devstral Small 25.05 for software engineering tasks and the powerful Mistral Medium 25.05 with multimodal capabilities), are built with openness at their core.

What makes Mistral stand out is their commitment to “open weights.” This means they don’t just give you access to the code; they give you the actual trained model’s “brain” — its parameters. This level of transparency is huge. It means you can inspect every nook and cranny, understand how the model learns, and even modify its core behavior. This contrasts with many “open core” models where the training data or key components remain proprietary.

Mistral’s success has shattered a powerful myth: that only companies with bottomless pockets and hundreds of thousands of GPUs can build cutting-edge foundation models. Mistral has shown that with smart architecture, innovative training techniques, and a focus on efficiency, a smaller, agile team can compete with the best of them. Their models are renowned for being incredibly fast and cost-effective to run, making them super attractive for developers and businesses who care about performance and their budget.

Their recent collaborations, like the partnership with Microsoft and HTX (Home Team Science and Technology Agency in Singapore) to fine-tune LLMs for public safety, demonstrate how even major players are recognizing the power and flexibility of integrating open-source models into critical infrastructure. Mistral isn’t just building models; they’re building an ethos.

The Open-Source Avalanche: How It’s Reshaping the AI Race

The impact of these open-source giants and the countless smaller projects they inspire is nothing short of transformative. It’s not just a trend; it’s a fundamental power shift in the global AI landscape.

1. AI for Everyone: The Ultimate Democratization

Imagine a world where powerful tools are only available to a select few. That’s how AI used to feel. Now, thanks to open source, that’s changing rapidly. A student in a remote Indian village with a decent internet connection can download and experiment with a model almost as powerful as the ones used by multinational corporations.

This isn’t just about charity; it’s about unlocking human potential. When more people have access to these tools, more diverse ideas emerge, more problems get solved, and innovation sparks in unexpected places. It means startups in emerging economies don’t have to spend a fortune on licensing fees, allowing them to innovate faster and bring local solutions to local problems. It’s truly making AI a public utility, rather than a secret weapon.

2. Innovation on Rocket Fuel: The Global Brain Trust

Think of a closed-source model as a brilliant scientist working alone in a lab. They might make incredible discoveries, but it’s just one mind. Now, imagine putting that scientist’s groundbreaking work out into the world, and letting thousands, even millions, of other brilliant minds inspect it, test it, improve upon it, and find new applications. That’s the power of open source.

When a model’s weights and code are openly available, the global developer community becomes a massive, distributed R&D department. Bugs are found faster, vulnerabilities are patched quicker, and new features or fine-tuning techniques emerge at an astonishing pace. This collective intelligence accelerates innovation beyond anything a single company could achieve, no matter how large. It’s a continuous, self-improving loop that pushes the boundaries of what AI can do every single day.

3. Bye-Bye, Big Bills: The Cost-Effectiveness Revolution

Training and running large AI models used to cost an arm and a leg. For many businesses, especially small to medium-sized enterprises (SMEs) or even large enterprises wary of massive cloud bills, this was a significant barrier. Open-source models slash these costs dramatically.

You can often run these models on much less expensive hardware, or even on your own servers, giving you more control and reducing your reliance on expensive cloud APIs. This financial freedom is a huge boon, allowing more companies to experiment with AI, integrate it into their operations, and build custom solutions without breaking the bank. It’s leading to a tangible return on investment (ROI) for businesses, making AI adoption a smart financial move.

4. Trust, But Verify: The Transparency Advantage

One of the biggest concerns with AI is its “black box” nature. How does it make decisions? Is it biased? Is it secure? With proprietary models, you often have to take the developer’s word for it. But with open source, you don’t.

Anyone can inspect the code, audit the training data (if available or reconstructed), and scrutinize the model’s behavior. This transparency is crucial for building trust, especially as AI is increasingly used in sensitive areas like healthcare, finance, and legal systems. It allows independent researchers to identify and mitigate biases, improve fairness, and strengthen security. It’s about accountability, ensuring AI serves humanity ethically and responsibly.

5. No More Lock-In: The Freedom to Customize

Imagine buying a car that you can never modify, upgrade, or even take to a different mechanic. That’s often the case with closed-source software. You’re locked into a vendor’s ecosystem, dependent on their updates and their terms.

Open-source AI breaks these chains. You can take the model, tweak it to your exact specifications, integrate it seamlessly with your existing systems, and build truly bespoke solutions. This flexibility is invaluable for businesses with unique needs or those operating in niche markets. It means you’re in control of your AI strategy, not beholden to a single provider. It truly enables a “build your own adventure” approach to AI.

6. The Power of Many: Community-Driven Improvement

Walk into the digital halls of platforms like Hugging Face, and you’ll witness the bustling heart of the open-source AI community. It’s a vibrant ecosystem where developers from every corner of the globe share models, contribute code, collaborate on projects, and collectively push the boundaries of what’s possible.

This collaborative spirit fosters rapid iteration and widespread knowledge sharing. Someone discovers a clever way to make a model more efficient? It’s shared. A new method for fine-tuning? It’s discussed and adopted. This constant flow of ideas and contributions ensures that open-source AI is continuously improving, often at a pace that proprietary models struggle to match. It’s truly a testament to the power of human collaboration.

Bumps in the Road: The Challenges We Still Face

While the open-source revolution is incredibly exciting, it’s not all sunshine and rainbows. There are real challenges we, as humanity, must address head-on:

The Shadowy Side: Safety and Misuse: With great power comes great responsibility, right? The widespread availability of powerful open-source AI models also increases the potential for misuse. Think about the rise of increasingly convincing “deepfakes” that can spread misinformation, or the possibility of malicious actors using AI to develop sophisticated cyberattacks. Ensuring that these powerful tools are used for good, and preventing their exploitation for harm, is a monumental task that requires global collaboration and constant vigilance. It’s an arms race where ethical guardrails are just as important as technical defenses.

The Wild West of Quality and Support: Unlike a commercial software product that comes with dedicated customer support and rigorous quality assurance, open-source projects rely on community contributions. This can sometimes mean varying levels of documentation, inconsistent updates, or a lack of formal, enterprise-grade support. For large organizations relying on AI for critical operations, this can be a hurdle. The community is working on solutions, like more robust open-source foundations and commercial support wrappers around open models, but it’s an ongoing effort.

Decoding “Open”: Licensing Nuances and IP Headaches: The term “open source” itself isn’t a single, clear-cut definition. There are various licenses (Apache, MIT, GPL, Meta’s LLaMA license, etc.), each with its own rules about usage, modification, and commercialization. Navigating these intellectual property rights and licensing models in a rapidly evolving field can be complex and confusing, leading to legal uncertainties. The community is still figuring out the best ways to balance openness with responsible innovation and commercial viability.

The Sustainability Question: Building and maintaining cutting-edge AI models, even open-source ones, requires significant resources — talent, computing power, and funding. While community contributions are immense, ensuring the long-term sustainability of major open-source AI projects is a challenge. Many rely on venture capital, grants, or hybrid business models (e.g., offering commercial services built on top of their open models). Finding stable and diverse funding streams is crucial for these projects to thrive.

The Hybrid Horizon: Where We’re Heading

As we push deeper into 2025 and beyond, it’s clear that the global AI landscape is unlikely to be dominated by a single approach. The “AI race” isn’t a zero-sum game. Instead, we’re witnessing the emergence of a hybrid AI ecosystem.

Proprietary models from the biggest players will likely continue to push the absolute bleeding edge, fueled by their vast research budgets and access to unparalleled computing resources. These models may excel in niche, highly complex, or highly sensitive applications where bespoke solutions and dedicated support are paramount.

However, open-source models are rapidly democratizing these advancements. They will continue to close the performance gap, becoming the engine of innovation for startups, researchers, and developers worldwide. They will drive localized solutions, foster diverse applications, and serve as a crucial check on the power of closed systems, demanding transparency and encouraging ethical development.

The future is one where businesses and innovators have a rich spectrum of choices. They can pick the best tool for the job, whether it’s a high-performance, proprietary model for a very specific task or a flexible, cost-effective open-source model that can be tailored and deployed anywhere. The competition between these two approaches will continue to accelerate innovation, driving better performance, lower costs, and more strategic flexibility for everyone.

This open-source revolution is a powerful testament to collective human ingenuity. It’s a bold statement that the future of AI belongs not just to a few, but to all of us. By embracing openness, collaboration, and shared knowledge, we are collectively building an AI-powered future that is more equitable, more innovative, and ultimately, more beneficial for all of humanity. And that, standing here in Fatehpur Range, feels like a future worth being excited about.

Conclusion

In this exhilarating, high-stakes game of AI chess, the open-source movement hasn’t just flipped the board; it’s invited millions of new players to the table, each armed with disruptive strategies and an insatiable hunger to innovate. Forget the hushed whispers of corporate labs and the guarded secrets of proprietary code. The future of AI is no longer a tightly held monopoly but a wild, collaborative frontier where the brightest minds from every corner of the globe are converging.

So, let the titans of tech guard their gilded towers; the real revolution is brewing in the open fields, where collective genius, unfettered and fierce, is forging an AI future that’s not just powerful, but truly unstoppable. This isn’t just about code or algorithms; it’s about a fundamental shift in power, a democratization that promises to unleash an explosion of creativity and problem-solving beyond our wildest dreams. The finish line? It’s not a single point, but a constantly expanding horizon, propelled forward by the relentless, open-source tide.

Sources: ashnik.com, pollingersocial.co.uk, linkedin.com

Authored By: Shorya Bisht