exhibit A: the suspect, age single digitsThis kid wanted to become everything.
Detective. Cricketer. Astronaut. Probably Batman at some point too.
Turns out he became a computer scientist instead.
I wasn't one of those kids who knew exactly what they wanted to become. Every few years I discovered something new, declared it my life's purpose, annoyed my parents for a few months, and moved on to the next obsession. Less of a career plan, more of a software update every two years.
scroll: the trail draws itself ↓
The first dream I remember was becoming a detective. Not the Sherlock-with-a-pipe kind, I wanted to join the CBI. Tiny me imagined solving impossible mysteries, catching criminals, and walking away from explosions that definitely weren't necessary.
Reality had other plans. I couldn't even find the ₹500 note that i lost in school. My detective career ended before it even began.
Naturally, I pivoted. If I couldn't solve crimes, I'd score centuries. Cricket became everything, I trained seriously, played professionally for two years, and represented the Under-14 side. It felt like I was one step away from being the next big thing.
Then I learned how Indian cricket selections work. Let's just say my optimism had a shorter career than my cover drive.
Then Christopher Nolan entered my life. Interstellar, The Martian. Suddenly I wanted to be an astronaut, space seemed amazing, and physics seemed manageable.
Then Class 12 happened. Newton and I developed creative differences. My dreams escaped Earth's gravity. My physics marks did not.
And like millions of Indian students before me, I eventually arrived at the most common plot twist in existence: Computer Science Engineering. Not because I had a ten-year roadmap, mostly because everyone around me seemed to agree that “CSE has scope.” Fair enough.
Before college even started, I joined a coaching institute to learn C. I still remember staring at the screen after i reversed an array. It looked... cool. But it didn't feel cool. Nobody wakes up thinking, “you know what would improve my day? Someone reversing an integer array.”
That's when it clicked: I didn't want to write code that just looked clever. I wanted to build things people could actually use.
So I opened YouTube, like every broke ahh engineering student with an internet connection, and discovered HTML and CSS. Huge shoutout to SuperSimpleDev, the man has probably taught half the internet to center a div. (The other half is probably still trying.)
One tutorial became twenty. Twenty became a hundred. HTML became CSS. CSS became JavaScript. JavaScript became React. React became Next.js. Tailwind appeared somewhere in the middle and suddenly everything looked prettier.
Around the same time, Python quietly entered the picture. At first it was just another language. Then it became machine learning. Then computer vision. Then automation. Then AI. Looking back, Python wasn't another tool, it was the beginning of a completely different chapter.
The tutorials ended and the building began. Real-time video apps. AI chatbots. Typing games. Hackathons. Google Solution Challenge. Things broke, got fixed, got shipped, and strangers on the internet actually used them.
Somewhere in that loop, “learning to code” quietly became “being a builder.”
Building led to questions that tutorials couldn't answer, and questions led to research. The first IEEE papers were on blockchain. But the deeper I went, the more one thread kept pulling at me: what happens when AI systems act on their own?
Blockchain turned into AI security. AI security turned into AI safety and multi-agent systems, AgentOps, evaluations, how autonomous agents fail, and how to catch them before they do. This is where I stopped asking “what can I build?” and started asking “how do we make sure what we build behaves?”
All that blockchain research eventually turned into something I didn't see coming, AI security and AI safety. Somewhere between paper after paper, one question hooked me and wouldn't let go: what happens when AI systems start acting on their own, and how do you even tell if one of them misbehaved?
No lab, no advisor, no funding, just a laptop and an unreasonable amount of stubbornness. Turns out that's enough to find a research area you actually want to stay in.
While all of that research was happening, I also went back to building, and made Sasana, a way to make AI agent logs tamper-evident so nobody can quietly edit or delete what an agent actually did.
Here's the GitHub: sahiee-dev/sasana
Next stop: UMass Amherst. Grad school, research, and the same kid who wanted to become everything, except now the obsessions finally compound instead of replacing each other.
The detective instincts became debugging. The cricket discipline became consistency. The astronaut dream became a fascination with systems bigger than myself. The journey is just getting started.