Now

Leading AI engineering at Autoscreen.ai — building LLM-powered evaluation systems that can assess candidates at scale without losing the nuance that humans bring to hiring decisions.

Exploring multi-agent architectures — specifically how to make agents that can decompose ambiguous problems, delegate to specialists, and synthesize results coherently.

Whether the right mental model for LLMs is closer to "very fast retrieval" or "genuine generalization" — and why the answer matters for how we build on top of them.

The gap between demos that impress and systems that ship. Most AI agent demos look magical until they hit the edge cases that real workloads produce constantly.