What I've Worked On
Most recently I was at Pixa AI as a Founding Engineer and Researcher. Co-built Luna, a SOTA speech model that understands meaning not just words. Also worked on neural audio codecs and an ASR model for how Indians actually speak, code-switching between Hindi and English. The ASR was also SOTA, beating Eleven Labs Scribe, Whisper, Deepgram, and Sarvam's model on FLEURS Hindi benchmarks. Built the data pipelines to make it all happen.
Now exploring agentic search and RL environments for niche dataset creation.
Before This
I first started exploring AI with something called Mindify, an edtech agent designed for low income families who couldn't afford traditional private tuition. I was invited to present this at the India Mobile Congress to Mr. Ashwini Vaishnaw and our honourable Prime Minister.
I skipped the traditional path. At 14, I built an NLP-based COVID tracking bot for Discord that went viral and served 500K+ users. Got the bot verified on Discord too (had to use my dad's ID since I was underage). Reverse-engineered Discord for hosting self-bots, got three IDs banned in the process, then moved on to learn Golang.
Built a web framework like Express but around 50x faster. Got over 150 stars organically and about five companies use it now. Moved to crypto for fun, came top 5 in my first hackathon.
During school, I spent most of my time skipping classes and building hardware projects in the labs. That hands-on tinkering helped me win both the Youth Ideathon (2022) and podium at the CBSE Science Exhibition (2022). Still the only person to achieve both. I later became a student evaluator for the CBSE Youth Ideathon.
I've won multiple premier hackathons, most recently Unfold '23 where I came 2nd as the youngest winner. Over the years I've received grants from the Government of India, CBSE, Solana, and Polygon for my projects.
Tried my hand at building a startup with Connect AI, a Jarvis-like interface that could convert thoughts into action. Went viral on Twitter, got reached out by Nexus VP, Sierra Ventures, Gradient Ventures, and Peak XV. Even got to visit Ola's office where they were doing human annotation, which was pretty cool. Had gaps in my knowledge when it came to business, couldn't really pull it off but got a lot of learnings.
Always open to conversations about AI research, interesting collaborations, or ambitious projects.