I'm a researcher based in Amsterdam, Netherlands. I mainly work on understanding and developing interpretable models that help in accelerating scientific research. My career goal is to better understand the nature of intelligence.
I am a Data Science lead at RespiQ, developing machine learning models and AI systems to build a smart healthcare monitoring device based on plasma emission.
What I'm doing now:
- Building autoencoder models for detecting unknown molecular compounds at RespiQ. Working on generalized detection across diverse chemical compounds without prior training on target molecules.
- Using machine learning models to understand the hardware limitations and improve the core plasma technology.
- Researching the viability of detecting circuits of syntactical and lexical systems in small language models. Small writing sample here
Things I've built:
Open Source
- SimpleStories Dataset - An open-source synthetic dataset for training small language models to advance mechanistic interpretability research. Accepted at NeurIPS 2025 Dataset and Benchmark track. (arXiv, Hugging Face, Github)
- OpenSim-RL Framework - Contributor to the Reinforcement Learning framework for musculoskeletal simulation research at University of Groningen. (GitHub, Paper)
Commercial Products
- Multi-agent LLM System - Fine-tuned large language model for medical query processing for healthcare data analytics platform deployed across multiple data warehouses for African healthcare ministries
- No code computer vision platform - Multi-camera with custom models for automated detections in Nvidia edge devices enabling partnerships with HP and Intel. (Case Study)
Other interests of mine include sci-fi novels, hiking, cooking, electronic music, and exploring the intersection of neuroscience, culture, and human potential.
I'm always interested in connecting with researchers and engineers working on interpretability, alignment, and related open-source projects. Feel free to reach out!