I've built production agentic systems at VoiceCaptures, including a multi-agent SMS outreach platform using the OpenAI Agents SDK and an AI voice receptionist for home service contractors. My systems connect LLMs, CRM workflows, SMS, retrieval, and human review into practical customer-facing products.
At the Canadian Institute for Health Information, I lead a RAG knowledge assistant initiative for internal documents and an ongoing responsible-AI and AI-safety workshop series. I also build distributed PySpark pipelines processing more than 500 million healthcare records annually and microsimulation models that improved long-term staffing forecast accuracy by 20%+.
In DeepAnalytic, I evaluated RAG over the Stanford Encyclopedia of Philosophy using section-aware retrieval, Pinecone, Cohere Rerank, and schema-enforced multi-query decomposition. I built an evaluation harness with golden answers, LLM-as-judge scoring, and independent human review—and let the results redirect the design when reranking lowered scores.
My PhD in formal logic and semantics shapes how I assess whether agents actually follow rules. Earlier, I used Python, SQL, machine learning, clustering, and regression across education, sales, and contract-evaluation problems.
