In this blog, we're going to talk about interests, passion, and curiosity. Every year when January rolls around, we tend to come up with resolutions and develop new interests. "Let's build something," we tell ourselves. "Let's develop some passion."
It may not always be impulsive. Most of the time, there's an offset period—as you continue to interact with certain content or specific things, you develop a feeling: "Oh, this makes sense to me," or "I find something relatable here." Then there's another way interests develop—you keep watching somebody doing reels or someone doing something specific, and you start thinking, "Let me try that too."
Sometimes it's family influence. Everyone in your family might have been inclined towards something. I've seen this with music—a father or grandfather carries an interest in music, and it naturally brings that same inclination to others in the family.
So imagine this as a reflection on how we develop interests in a specific field.
Take what's happening nowadays. Whoever is reading this blog might be deeply interested in artificial intelligence, machine learning, or data science. This is the highest interest I see among people right now. Or software development. These are all rank-one interests for a specific segment of people—the ones I'm trying to connect with. Everyone in this field would have started with the highest curiosity towards AI, data science, machine learning, business analytics, or parallel segments like design and user experience.
But where does it actually originate from?
For me? The interest wasn't there. Not at all.
2015: Just Get to Chennai
Back in 2018, during my fourth year, there was a campus placement drive. In Tamil Nadu colleges at that time, AI, machine learning, or data science weren't elaborately discussed topics. One or two guys might have known about it because their brothers or relatives worked in IT companies. Even I didn't know what these terms actually meant.
But at that moment, I was taking a subject called Applied Soft Computing. We were learning neural networks, and I saw it as a methodology you could use for control systems and building projects. "Oh, we need some good projects on our resume." And it sounded fascinating—you create neural networks that are similar to the human brain. There were genetic algorithms, metaheuristics, bio-inspired algorithms. "This is how an ant behaves, and we can use that strategy to develop an algorithm for optimizing control systems." Control systems like thermostats maintaining room temperature, chemical reactors, or self-driving cars. Sophisticated systems that need algorithms to maintain control.
That was the starting point, the origination point. And I was completely unaware of where I was heading.
Consider 2015. I chose my college just because I wanted to be in Chennai. That's it. I never had any expectations from St. Joseph's College of Engineering. I never had any context that this college would shape my thinking in any particular way. I didn't do well in my Plus Two, got around 190 engineering cutoff, and had very few choices. I was highly demotivated by advice like "Don't take computer science."
You might be wondering—no CS degree? Some of you might be thinking about IIT JEE preparations. The thing is, I'm coming from a background where we never had such exposure. No exposure, no idea. Just a simple goal: be in Chennai, get into a program where many people get jobs, so no relatives would laugh at me, and I could take care of my family.
2018: The Mu Sigma Interview
By 2018, I was having good grades. I was just recovering from post-surgery things. I had developed some understanding that Instrumentation and Control Engineering meant working on industrial systems. Unlike electrical or mechanical engineering, which had hype and recognition, ICE was available in very few colleges. So it was feasible to get a university rank. I'm being honest here—that was the thinking.
Then came the Mu Sigma interview. It wasn't like the usual TCS or CTS placement—aptitude tests, logical reasoning, English, resume analysis. No. This was fully managed through psychometric tests and mathematical questions.
I still remember—my ATM card was lost at that moment. I was so nervous, searching everywhere. I went into the interview panel, and the interviewer, Nagendra, asked me a lot of questions. I blabbered. I struggled to answer. He told me directly: "You are completely lost in your English. You don't talk well. You're not able to clearly communicate."
But he must have seen something—maybe through the group discussion scores or something in the interview. He said I needed to improve my English. He even found my ATM card and asked why I didn't inform anyone. I told him honestly: "I just want to get the job so I can manage my family, take care of my mom."
He said, "You didn't do well. Maybe you have to practice English every day. Just connect with me on Facebook." He actually connected with me on Facebook. Because back then, Facebook was the thing. It's been eight years now—2018 to 2026—and that conversation continued.
My English wasn't great, but I could communicate bits and pieces. I got an offer right then and there. Until January 7th, 2019, I wasn't sure what I was going to do. Some peers were getting selected at Zoho for six or seven lakhs. People were preparing for dual offers from CTS, Wipro, Cognizant. I didn't have much clarity about my career for the next few years. It was all about: "We got a job, we'll earn, we'll support the family."
2019: The Mu Sigma Immersion
January 2019, internship started. Four months, unpaid. I worked with a couple of teams—three or four mentors from each. The first project was with a pharmaceutical company—pharmaceutical forecasting, understanding how much demand a particular medicine would require in a specific region. The second was telecommunications—a customer churn problem. The beauty was how we were introduced to these problem statements.
There were courses taught by senior people—statistics, Python, Excel. There were exams with bands: orange, green, red. I remember crying to Manjari and Shruti, who were handling the MSU program: "I got orange, will I get rejected?"
Every Wednesday was Great Coding Day—8:30 AM to 7:30 or 8:30 PM, twelve solid hours. You'd be given a problem statement and dataset, and all MSU batches would participate. Every week for twelve weeks. It was a proper university experience. Most internships expect you to come in knowing something. Not Mu Sigma. They took raw crystals and shaped them over twelve weeks, giving you an entry-level view of how the company works.
Once I had a one-on-one with Sujeeth, he was our Incharge for Batch F3. I even showed my concerns about my orange bands, fear of getting fired. During that session, he asked me where I see myself in the next five years. Without doubt, I mentioned I wanted to become an Engagement Manager—which was the senior, important role at Mu Sigma. They drove the entire engagement, worked closely with clients, and were often at the client's place or onsite. To be honest, they were really good at management, consulting, and analytics. Most of the managers—we called them Apprentice Leaders—ended up doing MBA at top business schools around the world. Three out of six managers I worked directly with got into top MBA programs, and they were so brainy according to me.
Out of 600-800 people selected, only around 30 got internships. The company bet on people who didn't even know what data science was. They didn't use the term "data science"—they called it "decision sciences." A problem-solving company. They worked with hundreds of Fortune 500 clients, dealing with high-value, high-scale problems. Some engagements were worth millions.
That experience was rich. That's when the real feeling of interest started. Though I still didn't have control over it—it was just "this sounds good, this seems interesting."
2019-2022: The Golden Period
Three years passed. I worked with many clients and got a very good reputation. Mu Sigma was the only place where I received a handful of awards and appreciations. After 2022, the word "award" wasn't really in my vocabulary anymore. But even in June-July 2022, before leaving in October, I received an impact award for a project idea that converted into a full-time deal.
There was proven work with clients. One client from Japan specifically mentioned to a BU head: "We want to work with this particular person only." They visited us. I was taken to that client meeting. It was the golden period of my career—January 7, 2019, to October 8, 2022.
Then I left due to many reasons. As you learn, you develop your own motives—how you want to work, how you want to see projects. Sometimes contradictions come. Sometimes situations arise—mom wasn't doing well, financial constraints needed addressing. I'm not saying things went wrong on one side. It was just unexpected.
2022: Captain Fresh and Fishes
I got many interview offers and ended up joining Captain Fresh. Why was it interesting? They were dealing with fishes and shrimps. I saw it from a perspective that most people don't get—my peers were usually working on retail analytics or forecasting. At Mu Sigma, people interested in solving difficult problems got put on interesting, struggling problem statements. Sometimes we'd cry or fight. Lots of emotional love and hate with that ecosystem. Whatever people now talk about as "SF energy" or "craftiness"—we were doing that without realizing it.
When Parag interviewed me at Captain Fresh, I was fascinated. Why was he working on shrimps and fishes, on the ocean? It was a completely new dimension. For me, it was about exploring as much as possible. I wasn't just someone who knew DSA and coding—I could understand problems, think through solutions, and I had tools: statistics, machine learning.
At Captain Fresh, I did a lot of prototypes and testing—maybe 30% research. Research in the sense of bringing shape to a problem statement itself, understanding how it connects with customers. We tested, presented at conferences. It showed me how important it is for people in tech, especially in Indian companies, to do this kind of work. Sometimes I feel people in these novel problem spaces do better research work than university teams—they're just not publishing it.
The Shift Toward Living Systems
I was working on three different things: deep learning for semantic segmentation, causal inference methods to understand how pond conditions impacted yield, and product engineering prototypes. I developed an interest in coming up with new products—products that never existed, solving problems that didn't have clarity yet, requiring new algorithms and interfaces.
ChatGPT launched. I was doing NLP work even back in 2020—presenting to Japanese clients, teaching ten-day sessions from lemmatization and stemming to GPT-2, ULMFiT, ELMo, skip-gram, CBoW. Real problems on entity recognition, training models on my crying Lenovo laptop.
All the while, my exposures were across different things. Three months on market simulations, then fighting to transition to computer vision projects. I remember approaching an AL who was leaving: "You're leaving. I want the opportunity because I worked on neural networks in college." I fought for what sounded cool. I was going behind everything. But luckily, I was given opportunities to build, show, and deliver to clients.
Books and the Emergence of a Pattern
Parallelly, from December 2022 onwards—Christmas Eve, I remember—that's when I first bought books worth 3,000-4,000 rupees. I still remember Sabesan, my junior at Mu Sigma, now doing his masters in the Netherlands. We went to Crossword and bought books.
The first book was 5 AM Club. It was about routines, and I never actually followed waking up at 5 AM, but those case studies kindled something—how people behave, how things happen. I connected them with my most beloved project: agent-based modeling. Among all the projects I did at Mu Sigma—NLP, optimization, inventory simulation—that one stays closest to me.
One project was about consumer behavior. Another about mobility behavior—where people go, their paths, their directions. Another about how people drink Coca-Cola, choose Dr. Pepper, select Snapple, or drink Seven Up. Companies want to study this. Only about ten members had worked on agent-based modeling. No other companies were really doing it—a niche thing that Mu Sigma bet on.
During that time, Dhiraj used to talk about complexity science, and Zubin about agent-based modeling and decision-making. One direction was exploration of business problems. The other was Mu Sigma inculcating a specific thought process—a way of seeing problems through complexity science and decision science. I was so into it because it gave me a substrate: every problem looks similar when you think through it.
The Convergence
I started reading more—Thinking, Fast and Slow, Nudge, behavioral economics, David Deutsch, Gladwell. It was all around thinking. Thinking about humans. Humans existing in this world with its problems and patterns. I connected it with Mu Sigma's illustration that decision science encompasses behavioral economics, design thinking, statistical inference, machine learning, artificial intelligence.
We chase terminologies the world defines, but something had been continuously emerging in me. And I was naturally reading these books—not books on "how to do machine learning," but books about roots. Mathematics of machine learning. Vladimir Vapnik's statistical learning theory. Thinking, Fast and Slow—because the models we were building in early 2024 never had reasoning skills.
I connected it with everyday situations: why I'm not washing my bedsheets but ordering a new blanket. How society struggles, how companies struggle, how customers struggle. How engineers struggle with systems while also struggling with themselves. Studies around engineers and how they solve problems, how we can use their experience to bring it to AI. It was all about thinking—humans and algorithms.
Behavioral science, complexity science, and decision science—all intertwined by algorithms, data, and AI. I got it. But I was still restless, chasing multiple things. Sometimes I'd develop FOMO: "Why shouldn't I just train a simple sentence transformer and share it?" But there was no inner calling from that. I didn't feel I owned that problem. Just training a model—even though I've done it—felt purposeless unless it had a specific problem where I belonged.
Seven Years, One Pattern
Now, looking back from 2019 to 2025—seven years—I can see a pattern. It's beautiful how I've traveled, how I'm approaching convergence, how I'm relating whatever I learned from different domains.
Mu Sigma had this interdisciplinary property. It's not that I'm leaving what I learned in the past—I'm leveraging it, interconnecting. I'm reading Jeffrey West's Scale—how organisms, cities, and companies share common behaviors. This interdisciplinary interaction, where new thoughts bond when things meet.
Something has very naturally impacted me. Consider this my witness statement of how things have truly influenced me.
2026-2030: Where I Belong
Now, for the next five years—2026 to 2030—at this tipping point of AI and post-AGI, I feel confident and clear. I can see I belong to a problem space. It's not about where I can see opportunity, where I can get a job, what I can add to my resume. It's about where I belong and what I want to own.
This agency actually triggered when I decided to quit Informatica. One year ago, January 4th, I was finishing my application to UC Berkeley for my MS program. When friends asked why I was going for a masters, I told them literally: "I want to run the show." I even put it on my LinkedIn. People might find that weird, but LinkedIn is my daily journal, my vision board—where I share from experience than for engagement.
People might say it's a place of noise and cringe. That's fine. But my LinkedIn reflects how I see things. Now I truly say I belong to a specific space. I want ownership around this problem space. I want to contribute. These are the interesting problems to bet on for the longer run. I never optimized for "next job, next job"—otherwise it would have been a complete mess. But now, if you see the entire journey, it flows naturally.
Serendipity is the word that comes to mind. I'm recording this, and the word automatically appeared because if I reflect, it's happening. Self-organization.
I'm not preparing a script or asking ChatGPT to sound like someone else. It's my own inner voice retrieving seven, eight years of incidents, connecting them, weaving them into a fabric of reality. And with this fabric, I wrap myself and feel warm.
I feel I belong. I feel I own something. I feel the agency to operate in artificial intelligence.
From Getting a Job to Running the Show
Connect the dots: I joined college with nothing in my mind except getting a job. Should I have the same mentality at Columbia? Should I think, "All I want is a job"? No. What's the impact of that mindset?
This agency and confidence comes from knowing: this is my problem space, this is where I want to analyze, research, design, engineer, and feel confident. Continuously work on it. Keep moving. That's convergence.
In one LinkedIn post, I called it Agentic Decision Sciences. This idea—I pitched it to my 2015 self from 2025 May. Now it has very good clarity. The real interest is actually starting now. The interest isn't about fancy things, unique niches, or what only I've worked on. Now I see there exists a bigger problem, there exists scope, and my heart goes there. I feel I can do a very good job. I have learnings I can really use.
I don't say I'm "passionate" about it—the word "passion" has a nonsense trap. It's about ownership. So I'm going to build the talent, skills, and contributions. Let it lead me. Let it absorb me.
A Note on Agency
If you find my story relevant, you can find ideas about how to think about career, interest, and developing agency.
Andrej Karpathy mentioned that having agency higher than intelligence may not work. Someone has logic, someone has a cheat sheet, someone has a Twitter fad. We're all humans with our bandwidth and filtering capacity. You can't just read "agency is greater than intelligence" and suddenly become high-agency. It doesn't happen because you read a tweet or made it a resolution. It needs time to grow, to emerge from what you read, what you think, what you work on, how you keep moving around specific things.
Without clarity, without exposure, agency stays on paper, on metrics—not real.
If you find this relevant, if your trajectory is parallel to mine or cutting through it, I hope this gives you a different perception.
Thanks for reading.