Technical leader with 15+ years scaling distributed systems and deploying NLP/RL models in production. Specialist in Inference-Time Compute, Constrained Decoding, and Agentic Architectures. Combining deep research in reasoning with CTO-level execution in high-reliability enterprise environments.
Led two research-based ventures, Miras and Eveince. Shipped multiple AI products and published research on NLP, language models, and speech. Five patents in USPTO and EPO covering high-scale ML systems, in-memory computation, and error-free AI code generation.
Currently researching out-of-distribution generalization, test-time learning, and agent scaling. Building Sensi: Test-Time Induction Framework & Dynamic World Modeling.