I've spent over a decade in AI — building Bittensor subnets, shipping computer vision systems, competing against AlphaFold, fine-tuning one of the best tool-calling models in the world. I'm obsessed with these problems and I want to work on them.
Alignment with human values, and the race both to get there and away from there. Capability is compounding faster than safety is keeping up. I build adversarial evaluation systems where my job is to make honest participation the only viable strategy.
Recursive self-improvement through tooling. Models that call functions, chain actions, and modify their own behavior. I built one of the world's top-ranked tool-calling models and the infrastructure to evaluate what it does with that power.
Domain expert leading the rapid PoC team. Fine-tuning, agentic architecture, tool-use orchestration. Taking new AI capabilities from prototype to production fast.
World's top-ranked 8B tool-calling model. MCP server architecture for agentic systems. LLM fine-tuning, RAG pipelines, evaluation and sandboxing infrastructure. Incentive mechanism design across Subnets 20, 45, 64. 1000+ H200 GPU cluster via vLLM/Kubernetes.
Computer vision, predictive models from 5B+ sensor records, full AWS/Azure infrastructure.
Anomaly detection, OCR, pricing optimization. Taught Data Science.
First-author paper (IEEE/ACM). Competed against AlphaFold at CASP. BS in CS & Philosophy.