DataDrivenLab
AI and simulation for autonomous labs and materials — from semiconductors and photovoltaics to nanomaterials.
The company
DataDrivenLab Consulting builds physics-based simulators and their millisecond stand-ins, Bayesian inference that turns measurements into decisions, and the pipelines that run them.
Our mission
Lift your ambitions, then realize them. AI opens up the hard tails of engineering: thousands of weak signals combined into one answer, optimal designs found instead of guessed, and datasets large enough to settle the question. Our research runs with humans and LLMs in the loop: models propose, simulations check, people decide.
What we believe. Every step is traceable: a result is only finished when both a person and a machine can follow how it was made. Production and research should learn from themselves, so we turn them into self-improving information systems, where every measurement makes the next decision better. And the work should be a pleasure: our best results come when we enjoy finding out.
Our in-house R&D
We run our own research line because the hardest problems are the best test of our tools: every product is used there first, on our own data, before it reaches yours. Questions such as: what is the physics going on inside a perovskite solar cell while it degrades, and which measurements settle it; how to read the size and shape of nanoparticles from a single spectrum; and how far a hyperspectral camera's data can be compressed, at gigabytes per second, before it stops answering questions. Each began as curiosity and ended as software.
The founder
Sjoerd de Haan
Founder and principal scientist
- Education
- PhD Theoretical Physics, Groningen
MSc Mathematics, Cambridge
MSc Physics, Groningen - Training
- MIT Professional Education, Bioprocess Data Analytics & ML (Prof. Richard D. Braatz), 2024
- Stack
- Python, Julia, C++ · PyTorch, JAX · Bayesian ML, computer vision, Neural ODEs · Docker, Kubernetes, CI/CD, GPU infrastructure
Where I come from
I studied mathematics at Cambridge and did a PhD in theoretical physics. Then, from 2014 until now, I've been shipping ML to production: DataMeister in Tokyo, ING, TomTom, Huawei, a diagnostics start-up. A model on 100M+ phones, a validation design a lab still uses.
ML made huge steps in e-commerce. Science and engineering can gain far more from it, and have barely started. Since 2023 I've been working on that: simulators, autonomous labs, measurements that turn into decisions.
For heads of R&D and engineering
What we take care of
Direction
The technical roadmap for AI in your R&D and production: where it pays off a hundred times, where it does not, and in what order. Build-vs-buy and vendor calls you can defend to the board.
Delivery
Models, inference and dashboards that run in production, with the engineering to keep them there: cloud, pipelines, versioning, monitoring. Documentation investors and auditors can read.
Your team
Hiring your first data scientist, mentoring the ones you have, reviewing their work. The goal is a team that runs the loop after we leave.
Next step
Ask the bigger question.
What would you ask if every experiment had its answer by tomorrow? Start there. Your material, device or process follows. Picture what your science and engineering could reach when every measurement is answered the same day; the experiments follow from that question, and the loop that runs them stays yours.
Bring it to a free consultation. If we cannot help, we will say so.