At Intellum, I helped build the foundation of AI and data, taking the platform from no ML in production to an AI-first product. I built and shipped production features including course tutors, question-answering agents, an authoring assistant, and generative assessments.
I authored MCP servers that expose platform data to agents and built customer-facing MCP endpoints used with ChatGPT, Gemini, and Claude. I also created evaluation workflows using golden datasets and an LLM judge to catch regressions before release.
I helped build Intellum's first data platform, then designed and migrated its new architecture across event tracking, ingestion, data models, and data quality monitoring. Through data modeling and query optimization, I cut BigQuery costs by $100k a year.
Before Intellum, I worked on search and recommendation systems at Adevinta and OLX. At OLX, I designed its first recommendation system, which served more than 10 million users a day and drove over 20% of site contacts, and built an anti-spam and anti-fraud system that cut spam and fraud by 80%.

