Hi! I’m Ujjwal, a software engineer drawn to the frontier of technology. I try to work on the hardest problems I can find.

Past

In college I got interested in distributed systems and blockchain, and was the primary author of an NSF-funded research paper, SYMBIOTIC, on a permissionless IoT data-sharing network built on Ethereum.

After college, I spent two and a half years as a full-stack engineer at Kubera, a personal wealth management platform. My favorite work there was re-architecting Fast Forward, Kubera’s fully client-side financial projection engine, and leading Kubera’s AI integration end to end: implementing the newest OAuth standards, and building a custom GPT and later an MCP server.

Now

In February 2026, I left Kubera. I had built on top of frontier models, now I wanted to go a level deeper. I was inspired to work on a problem that first bugged me five years ago: explaining the key ideas in a chess position in natural language. I taught myself transformers from scratch and the basics of reinforcement learning, and got to work.

I am writing up the project as I go:

Contact

X
LinkedIn
Diagram comparing a student's sampled chess explanation token with the teacher's preferred next-token distribution
Distilling Reasoning: Exploring the Boundaries
I’m currently working towards training a task-specific LLM for chess reasoning: a model that can identify the key ideas in a chess position and explain them clearly in natural language. As a chess learner, my goal is to build something that I can learn from, rather than a model that simply outputs the best move. A natural first approach was knowledge distillation: using a stronger teacher LLM to transfer broad chess-reasoning ability to a smaller student. This was attractive because distillation can, in principle, recover much of a teacher’s capability with far less training data than conventional methods. To understand how effective and viable this approach might be for chess reasoning, I started with the literature.
2026/09/01 · Writings
Raw chess game data displayed beside an annotated chess book page
Synthetic Data for Chess Explanation: Building Human-Relevant Reasoning Traces
Introduction and Initial Approach I’m currently working on training an LLM that explains the key ideas in a chess position. In my previous post, I surveyed the literature on distilling chess explanations from a frontier model into a smaller one, and concluded that distillation was ill-suited for chess reasoning. The more important realization came from staring at the teacher’s outputs along the way.
2026/09/01 · Writings