In progress

AI Reliability for Systems

This series shares my research on exposing reliability failures in AI-driven systems and building practical solutions.
Thanks to all my amazing collaborators who make this work possible!

A free-form LLM iterates on arbitrary code and answers without evidence; with Constrained Creativity the LLM composes a DSL program run by a runtime engine and answers with evidence.

Making Agents More Reliable with Constrained Creativity

The counterintuitive path to more capable, lower-cost AI

Oct 2026· 5 min read
AI evolution turns program P into P′; AIChilles searches new inputs and exposes crashes, slowdowns, memory blowups, or P′ doing worse than P.

Are AI-evolved systems ready for primetime — Not yet!

Exposing the Achilles’ heel of AI-evolved systems

Sep 2026· 4 min read