At TwinWorks, I build the agentic layer behind a documentation and authoring product, including an AI copilot that inspects application state, searches documentation, and makes validated edits to procedure content.
I built a provider-neutral AI architecture spanning OpenAI, Gemini, AWS Bedrock, and self-hosted Ollama, making provider changes a configuration decision rather than a rewrite. I also created an internal LLM evaluation framework using golden datasets, LLM-as-judge scoring, and automated quality metrics.
I lead backend architecture for Rails and PostgreSQL APIs supporting 3D workflows, operating a multi-service AI stack with retrieval, embeddings, reranking, pgvector, Qdrant, Sidekiq, React, Docker Compose, and AWS. I drive testing and code-review standards, including taking one service from 0% to 80% coverage through a feature-flagged rollout.
Previously at Tap and Nawy, I built backend services, REST APIs, internal tools, and scalable data systems. At Tap, I reduced query response times by 40% and supported features used by 50K+ active users; through MishMash, I also build Rails APIs and Shopify/Liquid storefront components.
