# Bram Van Rompuy

> Senior Python engineer building LLM applications, MCP servers, agentic tooling, and developer workflows.

Website: https://bramvanrompuy.be/
Email: bram.van.rompuy@gmail.com
Location: Turnhout, Belgium
Service area: Belgium, Europe, Remote
Availability: Available for freelance LLM application, agentic tooling, and production Python work via own company.

## Direct answers
- Who is Bram Van Rompuy? Senior Python engineer building LLM applications, agentic tooling, and developer workflows. Freelance via my own company.
- What does Bram Van Rompuy build? LLM applications, retrieval systems, MCP servers, agent tool-use integrations, developer workflow tooling, and Python-heavy production systems.
- Where is Bram Van Rompuy based? Turnhout, Belgium, working with clients in Belgium, Europe, Remote.
- How can teams contact Bram Van Rompuy? Email bram.van.rompuy@gmail.com or book a call at https://cal.com/bram-van-rompuy-hhxc9y/30min.

## Best fit
- LLM application engineering: RAG, retrieval, prompt/context engineering, MCP servers, agent tool-use, evals.
- Agentic developer workflows: custom CLIs, harnesses, skills, and team adoption around AI coding assistants.
- Production Python: FastAPI backends, data pipelines, microservices, AWS, Docker, CI/CD.
- Technical advisory for founders or CTOs working on LLM tooling or agentic workflows.

## Less relevant
- Pure design-only work without engineering ownership.
- Commodity chatbot demos without production integration, evaluation, or retrieval quality work.
- Full-time roles that are not compatible with freelance engagement.

## Services
- LLM application engineering: RAG, vector databases, agent tool-use, MCP servers, prompt and context engineering, lightweight evals. End-to-end systems, not toy demos.
- Agentic developer workflows: Harnesses, custom CLIs, and skills around AI coding assistants so engineering teams can actually use them.
- Production Python and practical polyglot work: FastAPI backends, data pipelines, microservices. Python is the through-line; Go for single-binary CLIs, TypeScript where it earns its place, C++ at the engine layer.
- Technical advisory: Sounding-board for founders or early CTOs working on agentic workflows or LLM tooling. Low-volume, direct, practical.

## Selected work
- Maya MCP server: Open Model Context Protocol server that lets AI agents act inside Autodesk Maya: read scene state, run tools, verify their own work. Protocol-first, model-agnostic. Documented in a 3-part public case study including a real production-crash debugging walkthrough. Links: https://gimbalgoats.com/blog/what-is-maya-mcp, https://gimbalgoats.com/blog/inside-maya-mcp-architecture-examples, https://gimbalgoats.com/blog/debugging-a-maya-crash-with-maya-mcp
- Kaiwa: End-to-end RAG chatbot on Django and Celery. Ingests PDF, Word, and HTML; embeds with OpenAI; stores in Pinecone with per-user namespaces. Hybrid retrieval with Cohere reranking via LangChain.
- ActionRail: Viewport UI framework for Autodesk Maya. JSON-driven layout with a runtime Python API. 200+ commits, tests with coverage gating, ruff lint, structured documentation. Links: https://github.com/BramVR/ActionRail
- MayaCythonCli: CLI that compiles Python codebases into Cython extension packages. Target-based build matrix, GitHub Actions CI, and a bundled agent skill so AI coding clients can drive the build directly. Links: https://github.com/BramVR/MayaCythonCli
- goBankCli: Go CLI that connects to your local bank to read balances and transactions from the terminal. A local-first single-binary CLI with secrets via the OS keychain. Built as a practical stretch outside the Python comfort zone. Links: https://github.com/BramVR/goBankCli
- LEGO SKU-to-Maya pipeline: Python data pipeline on AWS that ingests 3D product models, processes them through Pixar USD, and stores queryable asset data in MongoDB. FastAPI backend on Docker served on-demand queries to production teams. Replaced manual handoffs with an automated flow.
- Agentic dev-workflow rollout: Internal harnesses, custom CLIs, and bundled agent skills around Claude Code and OpenAI Codex. Trained engineers on agentic workflows; built the tooling so the team could actually adopt them.

## Evidence
- Previous clients and teams: The LEGO Group, Scanline VFX, DNEG, Netflix, Gimbal Goats.
- 15 years of production Python work across VFX, product pipelines, developer tooling, and LLM infrastructure.
- Public case studies are available for the Maya MCP server.

## Writing
- What is Maya MCP?: Why Autodesk Maya needed its own Model Context Protocol server, and how the protocol-first design keeps it model-agnostic. https://gimbalgoats.com/blog/what-is-maya-mcp
- Inside Maya MCP: architecture and examples: How agents call tools, verify their own work, and what the tool surface looks like in practice. With live transcripts. https://gimbalgoats.com/blog/inside-maya-mcp-architecture-examples
- Debugging a Maya crash with Maya MCP: A real production-crash walkthrough where the agent identifies the offending node faster than a manual bisect. https://gimbalgoats.com/blog/debugging-a-maya-crash-with-maya-mcp
- A Hugo and Cloudflare Pages deployment stack: How this writing surface is deployed. Cheap, fast, repeatable. https://articles.bramvanrompuy.be

## FAQ
- What problems are the best fit? The best fit is production engineering around LLM applications, retrieval systems, MCP servers, agent tool-use, developer workflows, and Python-heavy backends. I am most useful when the work has to move from experiment to shipped system inside an existing product or team.
- Are you limited to Python? No. Python is the through-line, but I regularly work across TypeScript, Go, SQL, shell, Docker, AWS, CI/CD, vector databases, and LLM APIs. I can also operate near C++ or engine/tooling boundaries when the project needs it.
- Can you productionize an LLM prototype? Yes. A common engagement is taking a promising prototype and turning it into something testable, observable, maintainable, and connected to real data or workflows. That can include retrieval quality, tool boundaries, evals, deployment, cost control, and failure handling.
- Can you work inside our team? Yes. I am used to joining existing codebases, review habits, CI, stakeholder constraints, and production deadlines. I can own a narrow subsystem, pair with internal engineers, or act as the senior contributor who gets a first slice over the line.
- Do you advise or also build? Both, but the default is hands-on engineering. Advisory works best when you need technical direction, architecture review, or a second opinion; implementation works best when there is a concrete workflow, repo, integration, or launch target.
- What should we bring to a first call? Bring the problem, the current state, the constraints, and what a useful first result would look like. A repo shape, architecture sketch, failing workflow, sample data, or rough product goal is enough to decide whether the next step should be an audit, prototype, or implementation slice.

## Contact
- Email: bram.van.rompuy@gmail.com
- Cal.com: https://cal.com/bram-van-rompuy-hhxc9y/30min
- LinkedIn: https://linkedin.com/in/bram-van-rompuy
- GitHub: https://github.com/BramVR