AI Systems Architect

AI Systems Architect specializing in building inference infrastructure, agentic pipelines, and production ML stacks. Proven track record includes being Employee #1 at Refact.ai, co-building a SOTA HumanEval dataset, developing enterprise LLM inference backends, RAG systems, and SWE-bench-compatible agent loops. Also served as Interim Head of AI, shipping full air-gapped ML stacks (training, inference, synthetic data, fine-tuned models) in under 6 months. Currently working on a sub-400ms voice-to-voice cascade on consumer GPUs and a zero-dependency MCP server in pure Python. Operates as Nautiloid Protocol LLC, offering clean B2B engagement for AI infrastructure contracts where output is judged by systems shipped.