Jay RodriguezBuilding and evaluating LLM systems for regulated, multilingual domains.

I build and evaluate LLM systems for industries where a wrong answer has consequences: customs and trade compliance, healthcare, cross-border payments. At diGenius AI I shipped a zero-to-one enterprise AI platform for multi-unit restaurant operators: retrieval-augmented generation over 3,000+ corporate records on LangChain and Azure AI Foundry, four production apps, 500+ automated deployments, and CI/CD security gates that block vulnerable dependencies before release. At Puente AI, which I founded, I run a deployed RAG system (LanceDB, Google Cloud Run) answering trade-compliance questions for the US-Latin America corridor. It doubles as the testbed for my research question: how LLM compliance behavior degrades across English, Spanish, and Portuguese in regulated settings, which I'm pursuing this fall through the Algoverse AI Research program.

I came to machine learning from the customer's side. A degree in supply-chain management, then six years selling compliance-driven products to restaurants, law firms, healthcare providers, and financial institutions, the people whose problems I now build for. That's why my research question exists.