projects

Built and shipped a full-stack health product (React, FastAPI, Supabase) on GCP: a client app and a practitioner report review/approval dashboard driving a multi-agent deep research pipeline (pydantic-ai) that runs as queue-based background jobs, generating personalized care plans via hybrid RAG (Vertex AI Discovery Engine + re-ranking), with Postgres RLS and Terraform-managed Cloud Run.

Full-stack internal observability dashboard surfacing complex Databricks metadata to engineering teams, with a Kubernetes-based ingestion framework feeding the Cortex internal developer platform, cutting cloud spend by $1.5M+ annually for clients like ASMLGlobal leader in semiconductor lithography systems, enabling the production of advanced microchips.↗ Visit website and SephoraMultinational retailer of personal care and beauty products with 2,700+ stores worldwide.↗ Visit website.

Proof-of-concept for a portable lakehouse solution using DuckDB to modernize legacy data architectures. Integrated with Unity Catalog for governed, high-performance analytics.

Led the industrialization of a product NLP classification model at PostNLThe national postal service of the Netherlands, delivering 1M+ parcels daily.↗ Visit website, building the backend ML and data infrastructure from scratch — €1M+ savings (Innovation of the Year award).