Quantitative Research + Production ML engineer based in Amsterdam.
I build end-to-end ML systems and research workflows: data pipelines, feature engineering, validation, training and evaluation gates, deployment, monitoring and retraining, and reproducible release processes.
Start here: punkpredictor-case-study (sanitized architecture, operations, and evaluation workflow for a live production system).
Some of my strongest production work is in private repositories tied to active infrastructure and client systems; the case study above is the best public view of my architecture/process approach.
- Built and operate a production valuation system for CryptoPunks with scheduled inference, policy-driven retraining, and gated promotion/rollback.
- Live valuation quality (as of Feb 2026): 2,807 sales; MAPE 6.92%; median 4.68%; signed bias -0.80%.
- Forward/live strategy validation vs passive baseline: Sharpe 2.23 vs 1.57; max drawdown 16.32% vs 41.02%.
- Fine-tuned and deployed an open-source LLM on ~30k real chat conversations (QLoRA + production optimizations).
- Impact: lead conversion +20.3pp and P95 latency -85% to 8.9s.
- Built an internal operations platform with React/TypeScript, FastAPI, PostgreSQL, and Docker.
- Quant research and statistical validation
- Production ML systems and MLOps
- Forecasting and pricing models
- Applied ML for business workflows
- Crypto and on-chain data products
- punkpredictor-case-study - Sanitized production ML + quant research case study (architecture, release gates, monitoring, retraining policy).
- verifiable-ml-nft-metrics - On-chain verifiable ML metrics using sparse Merkle proofs.
- Predict-CryptoPunks-Price - Hackathon project (1st prize) for CryptoPunks price prediction and verifiable ML experiments.
- crypto-agent-memecoin-prototype - Event project demo combining market data collection, modeling, and an agent workflow (participant project; no prize).
- defi-loan-liquidation-prediction - Event project for DeFi loan liquidation ratio prediction (3rd place).
- ethdam-erc20-success-prediction - Event project for ERC-20 success prediction from contract features (Trail of Bits track prize).
- Springer (InfoSec) - Co-authored research on dark web market intelligence and credential exposure analysis.
- IEEE (Blockchain/Networking) - Co-authored work on a browser extension for crypto donations.
Python, TypeScript, SQL, PyTorch, TensorFlow, Hugging Face, FastAPI, PostgreSQL, Docker, AWS, Vercel, Heroku