Freelance AI Engineer in Amsterdam — Jeroen Overschie
Hi!! Super nice to meet you – I'm Jeroen. I've worked in the software domain for over a decade and in Data & AI for about five years. First as a Data Scientist, then ML Engineer, now AI Engineer. Before that I spent about seven years doing freelance web development alongside my studies. From September 2026 I am for freelance work again, based in Amsterdam, working in Dutch or English.
My work has tracked the field closely. I have done classical machine learning and MLOps at TMNL and Vattenfall, then a run of RAG and GenAI projects at NPO, Klaverblad, ING and DataSnipper. More recentlyAgentic systems and search: Agentspace at BCG, an Agentic ELT pipeline built with Agent Development Kit (ADK), and a hybrid search engine for FD Mediagroep that now serves millions of requests a year. Specializing on Google Cloud Platform, and specialized with getting something running in production rather than a recommendation. Find more elaborate project descriptions on my LinkedIn.
Alongside client work I've given over ten conference talks at PyData, PyCon and others, taught MLOps and LLM trainings at companies like Rabobank, Bol and ING, and written a fair number of blog posts. I very much enjoy sharing the story. My research and side projects are documented here too, and my Open Source contributor activity can be found on GitHub.
What I can do for you
- Freelance AI Engineer: production AI agents, end to end - multi-agent orchestration with Google's Agent Development Kit, deployment on Vertex AI Agent Engine (now Gemini Enterprise Agent Platform), MCP and tool integration into existing systems, plus the reliability testing, observability and evaluation needed to actually operate them.
Example: I founded and shipped Bonbonnetje, an expense agent now running in production across Xebia and handling hundreds of submissions a month (see the LinkedIn post); and designed an agentic self-serve ELT system for Hitachi using dbt-mcp and ADK. - Freelance GenAI Engineer: RAG and search systems, structured data extraction, and the evaluation that tells you whether they're working.
Example: hybrid keyword-and-vector search for FD Mediagroep on GCP - Pub/Sub ingestion, Cloud Run, Weaviate, Gemini embeddings - taken from architecture to production as lead engineer. - Freelance Machine Learning Engineer: models and the engineering around them - Vertex AI pipelines, CI/CD, monitoring, model registries, and the practices that keep models maintainable after launch. Example: predictive maintenance models and an MLOps platform for Vattenfall's European wind parks; an anti-money-laundering model taken to production at TMNL on distributed Spark pipelines; an MLOps platform review for de Bijenkorf.
- Freelance Full Stack Engineer: Python backends, web frontends, chat surfaces like Slack and Teams where agents increasingly live, and Google Cloud infrastructure managed with Terraform. Example: seven years of freelance web development in Ember.js and React before I moved into ML, and end-to-end ownership of the FD search platform including IaC with Terraform and Atlantis and CI/CD on GitLab.
Curious to collaborate? Getting in touch
I'm available from September 2026, based in Amsterdam, remote or on-site in the Randstad. If you've got an agent that needs to reach production, a RAG system behaving inconsistently, or an ML setup that needs a review, book a slot and tell me about it.
