Hi, I'm Harsh. Let me just be honest about what I do: mostly machine learning research, but I'm curious about a lot more than that.
Right now I'm an AI Intern at Rowboat Labs (YC S24) and an AI Resident at Lossfunk in Bangalore. I got into all this through a friend who showed me CNNs, drifted from fine-tuning LLMs into actual research, and stuck around for the unknown parts, the ones where there isn't a tutorial yet.
These days I think a lot about representation learning in sequence models, mechanistic interpretability, and state space architectures like Mamba. Longer term I care about world models and where neuroscience meets AI, and I want to document that journey as it unfolds.
But I'm not only an ML person, and I didn't want a page that pretends I am. I make music, take photos, read too much, play an unreasonable amount of Minecraft, and fall down rabbit holes about philosophy, consciousness, movies, and self-improvement. If something's interesting enough to keep me up, it's fair game.
That's mostly what the blog is for. I write to think clearly, in public, and I'd rather be honestly confused than confidently wrong. Read it here.
Say hi on twitter, or email me at harshvardhanvatsa@gmail.com if you want to build something together.