At Logica, I spearheaded multi-agent systems for autonomous infrastructure management with LangGraph and PydanticAI. I built production LLM integrations and RAG pipelines that help agents reason over operational data and make infrastructure decisions.
I also developed machine learning pipelines for predictive resource scaling and GPU configuration optimization, reducing infrastructure over-provisioning by up to 25%. My work included connecting AI recommendations to developer workflows and monitoring quality, latency, and cost in production.
At Trunk Tools, I led development of an AI-powered knowledge graph and document understanding system for construction project intelligence. I built RAG and agent features for searching project documents, identifying risks, and recommending workflows; retrieval improvements raised answer precision from 71% to 89%.
At Quizlet, I owned a personalized learning engine and built conversational tutoring agents that reduced student help-desk volume by 40%. Earlier at Coveo, I built search and recommendation systems, contributing to improvements in click-through rate and average order value.
