Almere · NL
8+ years shipping production systems end-to-end, from distributed data infrastructure to multi-agent LLM platforms and full-stack microservices.
A geospatial similarity engine that learns the structural fingerprint of a place from OpenStreetMap features and returns 'what other places are like this?' across multiple countries. Custom-trained vision-transformer embeddings, pgvector, MCP server, agent-driven map UI. Built end-to-end.
A multi-tenant data-quality platform: a Python SDK that runs inside each team's Databricks workspace, a FastAPI + Streamlit control plane, a fact/dim PostgreSQL store for results, threshold-based Slack alerting, and a no-code UI for non-engineers. Built on Great Expectations with a custom-expectations API for team-contributed rules.
An AI-powered Slack bot that reads a data-collection-plan document on Confluence, cross-references it against the live event registry and metrics catalog, optionally analyses a GitHub PR alongside, and posts a structured pass/fail review back into the thread. Multi-agent workflow with Agno + ChromaDB vector search + deterministic result caching.
Headless plugin giving any Vercel-AI-SDK-compatible LLM (Anthropic, OpenAI, Azure, Google) conversational control over a TomTom map. Contributed to design and implementation.
Two production Model Context Protocol servers giving AI agents structured access to TomTom APIs. Both open source, both authored end-to-end (TypeScript, design + impl + docs).
Production multi-agent system for traffic analytics, demoed by Product Management at CES 2026 Las Vegas. Five specialised agents (routing, route monitoring, junction analytics, area analysis, traffic volumes) orchestrated through Agno's Team abstraction, served over AG-UI v1.0 with persistent chat memory.
A production agentic navigation system that turns natural-language queries ('find me a quiet café between my next two meetings') into structured plans grounded in real-time location data. Full-stack, React/Next.js + Python FastAPI + Azure OpenAI, with a RAG pipeline for semantic POI search and autonomous multi-step execution planning.
Refactored a vehicle-telemetry processing pipeline. Transforms raw nested JSON / protobuf from millions of in-car navigation clients into clean Delta tables that power navigation-quality dashboards. Designed for query simplicity: PMs answer 'route success rate in country X this week' with a single SELECT.
Refactored the inconsistent set of ETL pipelines feeding TomTom's API analytics into a single OOP template, distributed as an internal Python package. Engineers subclass the base, get live Azure Data Explorer connections for free, and only write the business logic in extract / transform / load. Result: consistent, version-controlled pipelines across volume, response-time, and error-rate use cases.
REST API layer that powers the analytics dashboards on developer.tomtom.com. Sits on top of an Azure Data Explorer (Kusto) backend that ingests every API call across TomTom's developer products. Volume reports, response-time percentiles, error-rate breakdowns, and per-product usage all flow through this layer.
A PySpark-based Python package abstracting the I/O patterns between Azure Databricks and the half-dozen sinks an enterprise pipeline typically needs. Built during the ABN AMRO Bank engagement and open-sourced internally there.
Lead engineer on an enterprise data-quality platform with a pluggable connector framework for profiling, validation, and reconciliation across heterogeneous sources. Adopted across multiple internal customer engagements.
posts · photos · videos · notes · games →
Cosine top-K is doing the right math but the wrong job. Here is the gotcha, the fix (MMR), and a second fix on top of that (geographic deduplication) that production systems actually need.
Lessons from shipping a multi-tenant data-quality platform across many Databricks workspaces. Why centralising contracts beats centralising data, and how to build the SDK + control plane + fact/dim store pattern from scratch.