Michail Mitsakis

Materials Scientist · electrochemistry, materials informatics, industrial decarbonization

prof_pic.jpg

MSc Engineering - Physics & Nanotechnology

BSc Physics - Materials Science

I’m a motivated materials scientist, passionate about decarbonization, sustainable energy solutions and accelerating R&D for low-carbon tech through electrochemistry, Power-to-X and AI/ML for materials discovery & design.

During my studies, I obtained experience in cleanroom and wetlab experimentation and manufacturing, as well as DFT modelling, chemical descriptors, materials screening and optimization. In particular, during my MSc thesis I optimized electrodeposition parameters for hydrogen evolution reaction (HER) catalysts. Specifically, I utilized multi-objective Bayesian Optimization (BO) to maximize the activity and stability of electrodeposited Ni-W (nickel-tungsten) alloys under various pre-cursor concentrations, current densities, temperatures, deposition times and pHs by minimizing their overpotential and overpotential difference (obtained before and after stress-testing the catalysts under various conditions).

After finishing my graduate studies, I continued that work out of deep interest in the field, and developed the same BO method through a more simplified, modular and easier-to-use platform for scientific experimentation, Ax.

Today I work at a business and engineering consultancy on deep-tech and EU-funded programmes: CCS & hydrogen technologies, manufacturing data spaces, and shipyard digitalisation. That means proposal development, technoeconomic scoping, project management, and working directly with industrial and research partners.

Alongside that, in my free time I have been diving deeper into the latest research in AI applications for materials discovery and optimization, such as self-driving labs (SDLs), machine-learning interatomic potentials (MLIPs), advanced BO & LLM use, and more, while participating in hackathons and prototyping certain apps to further advance my understanding of the field. For example:

  • catalyst-kg-agent — a knowledge-graph-grounded, cost-aware multi-agent system that decides when a cheap database lookup is enough and when an expensive simulation is justified, within a simulated SDL-like automated workflow; built to understand the complexities involved in automated discovery
  • notion-second-brain — a fully local RAG agent with hybrid retrieval, reranking, and an evaluation harness; for talking to my Notion notes offline
  • baybe-corrosion-inhibitors — Bayesian optimisation with BayBE to screen small-molecule corrosion inhibitors for aluminium alloys, comparing molecular encodings against random search and testing transfer learning between alloys; a hackathon project exploring how far BO can cut the number of real experiments
  • camel-rag — a retrieval-augmented LLM that answers natural-language questions about catalyst adsorption energies, grounded in Open Catalyst DFT data and citing the records it draws on; a hackathon project on making large catalysis datasets queryable without losing traceability

At the moment, I’m actively looking for an industrial PhD or an applied research role in AI for materials science or manufacturing in Europe.

You can find takeaways from my readings, research and projects in writing.

I love hearing from people, so please reach out!

selected publications

  1. arXiv
    From Knowledge to Action: Outcomes of the 2025 Large Language Model (LLM) Hackathon for Applications in Materials Science and Chemistry
    Aritra Roy, Kevin Shen, Andrew MacBride, and 350 more authors
    arXiv preprint arXiv:2605.03205, May 2026
  2. ChemRxiv
    Bayesian Optimization Hackathon for Chemistry and Materials
    Sterling Baird, Mehrad Ansari, Zartashia Afzal, and 122 more authors
    ChemRxiv, Jun 2025