I build AI systems designed to operate under real-world constraints. My work centers on real-time inference, embedding and retrieval pipelines, and agentic systems built close to the execution layer, where latency, memory, and reliability matter.

I work end to end, from low-level model execution and interfaces to orchestration, streaming, and system behavior in production environments. Most projects here are systems in motion rather than polished demos.

I am particularly interested in decision-critical and automation-heavy domains, where correctness and failure modes are first-class concerns.

• Agentic and tool-driven AI systems
• Real-time speech and multimodal pipelines
• High-performance embedding and retrieval
• Perception and computer vision systems
• AI infrastructure and execution layers

An active systems workspace.

• End-to-end systems over isolated examples
• Infrastructure- and execution-focused code
• Designs shaped by deployment constraints