Zenari — Amsterdam
Now I do it for companies. AI strategy and the systems that follow from it — scoped, built and handed over by one person with a research background and a production habit.
PhD Evolutionary robotics, VU Amsterdam 12 peer-reviewed papers 14 yrs shipping software
(1+λ) evolution strategy fitting a gait controller to a target gen 0000 σ 0.340 rmse 0.000
What I do
Most AI programmes fail on the seam between the people who sold the idea and the people who have to build it. Here there is no seam.
You have a problem and a lot of noise around it. I work out whether AI is the right instrument, what it would cost to run, and what breaks first — then hand you a plan your own engineers can execute.
No handover to a delivery pod. The same person who wrote the architecture writes the code, ships it, and is there when it misbehaves at 3am.
Selected work
Client names are withheld where the engagement was under NDA. Everything else — the problem, the method, the outcome — is as it happened.
Approach
Evolutionary search works because most candidates are discarded fast. The same is true of AI portfolios: the value is in how quickly you can tell a promising use case from an expensive one.
Two weeks, at most. What is the decision being made, what does it cost today, and what data actually exists to support it.
The smallest experiment that can kill the idea. Cheap failure early is the whole point — most use cases should die here.
Production from the first commit. Evaluation harness, observability and rollback before features.
Your team runs it. Documentation, runbooks, and a working knowledge transfer — not a consultant on retainer forever.
Who you get
PhD, Evolutionary Robotics — Vrije Universiteit Amsterdam
My doctorate was in evolutionary robotics at the Vrije Universiteit Amsterdam: robots whose bodies and brains both evolve, and which have to learn to walk in the body they were born with. It is a field where nothing works the first time and the only honest measure is whether the thing moved.
That is a useful habit to bring to enterprise AI. I have since built agent platforms in Rust, computer-vision pipelines that had to run on a CPU at an airport gate, forecasting and battery-optimisation models for the Dutch grid, and AI portfolios for manufacturers who needed to know which twenty of their ninety ideas were real.
Zenari is my own practice. You get me, from the first scoping call to the handover — not a proposal written by one team and delivered by another.
Track record
Research
Twelve papers on evolutionary robotics, modular-robot learning and embedded systems. The full list, with DOIs and PDFs, is on the research page.
Lamarckian Evolution of Simulated Modular Robots
M. Jelisavcic, K. Glette, E. Haasdijk, A. E. Eiben · Frontiers in Robotics and AI, 2019
Morphological Attractors in Darwinian and Lamarckian Evolutionary Robot Systems
M. Jelisavcic, K. Miras, A. E. Eiben · IEEE Symposium Series on Computational Intelligence (SSCI), 2018
Analysing the Relative Importance of Robot Brains and Bodies
M. Jelisavcic, D. M. Roijers, A. E. Eiben · The 2018 Conference on Artificial Life (ALIFE), 2018
Directed Locomotion for Modular Robots with Evolvable Morphologies
G. Lan, M. Jelisavcic, D. M. Roijers, E. Haasdijk, A. E. Eiben · Parallel Problem Solving from Nature (PPSN XV), 2018
Contact
A short description of the problem is enough to start. If it is not something I should be doing, I will say so and point you somewhere better.