Research

We study the hard problems in AI. Then we turn the answers into products people rely on.

• 12 Nov 2026State-Aware Adaptive Recovery for Reliable Tool-Using Large Language Model Agents Under Dynamic Execution ConditionsDeveloping an adaptive recovery framework that enables LLM agents to detect execution failures, understand their current state, and select appropriate recovery strategies to complete tasks reliably under dynamic conditions.In progress