How I Went from a Small Town Kid to Working on Cutting-Edge AI

A compilation of learnings from wearing multiple hats in the chaotic environment of early-stage startups — the good, the bad, and the wild ride of pursuing this craft.

---

Part 1: Small Town Roots, Kota & Dropping Out

So here’s my story — I grew up in a small town in Rajasthan, raised by my kickass single mom and elder sister. School was my thing, and when I hit 11th grade, I fell hard for C++ programming. Like every other Indian kid with decent grades, I ended up at Allen Institute Kota for IIT-JEE prep. Spoiler alert: The pressure cooker environment there? Yeah, not my scene.

Fig. 1 — Mumma bought me a Pentium Dual Core PC after 10th standard, played a lot of CS/CoD and coded in C++

Landed at BIT Bangalore for telecommunications engineering instead. Failed a year — but honestly? That turned out to be a blessing in disguise. Instead of moping around, I hit the gym and dove deep into programming classes. Best decision ever.

My first real break came with an internship at this startup called Mobisy. Got my hands dirty with their Bizom app — hybrid mobile development with Ionic Cordova. Even worked on this cool computer vision project counting liquor bottles in store photos. Not bad for a first gig, right?

Then came the big leap — dropped out of engineering (sorry mom!) and jumped into Signzy, this early-stage startup founded by some IIT folks. They’d just raised angel funding and were figuring out digital KYC stuff. Started as an intern, worked my way up to full stack dev, got to work directly with the CTO and help the CEO with product management. Even represented us at some fintech accelerators. Spent a solid 1.5 years there, soaking it all in.

Fig. 2 — Engineer #1 at Signzy, early days — failed, learnt, enjoyed

Fig. 3 — Office boy @ Signzy performing his duties in our first ever post-funding Goa offsite

---

Part 2: Protodit & Finding Founder-Market Fit

The entrepreneurial bug bit me hard after that. Had this idea called Protodit — wanted to automate the hell out of financial statutory audit work. Did the whole user interview thing, tried understanding the market. Joined another fintech startup in social entrepreneurship meanwhile (Artoo), where the founder was cool enough to let me explore my side project. Built their web dashboard in Ember.js, played around with IndiaStack APIs. Then Artoo couldn’t raise their next round, and we all got laid off. Classic startup story, huh?

Tried pushing Protodit further — more interviews, tried getting ex-colleagues to join as cofounders, pitched to angels for feedback. But here’s what I learned — sometimes you’re just not the right person to solve a problem, even if it’s a real one. Founder-market fit is a thing, folks.

Fig. 4 — Nothing is going right… so I bought myself a Les Paul

Next up: almost became a CTO for an employee experience startup. Built them an MVP in record time, got some LOIs through demos. But when it came to equity talks? Founder pulled a fast one — wanted me as a “consultant” with less than 10% equity. Yeah, no thanks. Walked away from that one.

---

Part 3: Vectorly, Video Neural Shaders & Hopin Acqui-Hire

Then came Vectorly — this cool video compression startup out of MIT’s entrepreneurship cell (MIT Sandbox). Three founders, lots of pivots trying to find product-market fit.

Fig. 5 — COVID lockdown scenes: hacking an SVG-based video player with an in-house developed video codec

Fig. 6 — Weekend trips with team to Hampi & Coorg when Sam was visiting us

We engineered real-time browser-native video neural upscaling using WebGL shaders and WASM. Eventually, Vectorly got acqui-hired by Hopin (the massive COVID-era virtual events hypergrowth rocketship).

At Hopin and later StreamYard, I finally got to play with the really fun stuff — full-stack ML, building datasets, writing automated tests, doing internal research, and building WebRTC/WebCodecs broadcast infrastructure handling millions of live streams. Built SDK integrations, POCs for AI features in video conferencing, and created a smart clip detection algorithm using LLMs for podcast content.

Fig. 7 — Visual representation of my 5-year startup wild ride: Vectorly (0→1) → Hopin (10→100) → StreamYard

---

Part 4: Sabbatical, Digital Nomading & Building with Purpose

After the intense hypergrowth cycle and post-COVID shifts across the tech industry, instead of jumping right back into another corporate role, I made a conscious choice: take time off from the formal 9-to-5, learn to slow down, and spend time understanding what building with purpose actually looks like.

Fig. 8 — Digital nomading in South East Asia for a year

Fig. 9 — Fav Indian restaurant in Koh Phangan and my fav Airbnb with coworking in Bangkok

Fig. 10 — 7 countries, 7 months of nomading, countless memories — wild 2023

Fig. 11 — The luxurious POV of witnessing sunrise and sunset in your slow life era 🕊️

Know what’s funny? That kid struggling in Kota would never believe where tech would take him. But here we are — building Katana, MotionVector, and Lookaround in the open. And honestly? I’m just getting started.

---

Also published on Substack and Medium.