About
I'm an AI researcher building Unintellect. My goal is to make AI useful across sectors and improve the work people do every day.
A lot of generative AI looks impressive in a demo but never changes how work gets done. I care about the step after the demo: finding a real problem, building a system people can rely on, and measuring whether it actually helps.
My work spans model architectures, efficient inference, agent evaluation, and production AI systems. Through Unintellect, I put that research into practice on real problems.
Research Interests
- Applied AI: Building and evaluating systems around real workflows, with results people can measure and use.
- Small language models (SLMs): Efficient architectures, fine-tuning, quantization, and models specialized for specific tasks.
- Self-improving agents: Autonomous self-refinement via harness & weight updates, test-time reinforcement learning, and evolving agent scaffolding.
- Agent evaluation: Adaptive proxy evaluation, MLE-bench, and benchmarks for agents that build agents (AIE-Bench).
- LLM systems & inference: Scalable inference pipelines and production-ready agent infrastructure.
- Structured reasoning: Socrates, structured questioning, and long chain-of-thought (LongCoT) reasoning in AI research agents.
What I'm working toward
I want AI to be adopted wherever it can make a meaningful difference: helping teams spend less time on repetitive work, make better decisions, and serve people more effectively. The measure of success is what changes for the people using it.
Unintellect is where I'm putting that goal into practice, starting with specialized models for business workflows.
Feel free to reach out: [email protected]