Mojmir Mutny

Experiment design, Bayesian optimization or Active learning -- all under one umbrella. The advent of self-driving labs is here. We need strategies to implement automatic information gathering! ML models are only as informed as the data they are trained on.

ML and automation can take protein engineering to a whole new level. In a really exciting collaboration, we have explored new strategies to navigate large sequence-activity landscapes in an efficient yet thorough manner. pubs.acs.org/doi/10.1021/ac…

Meet Heisenberg, the software agent that cloud controls chemputers to make molecules using our ChemputerAGI, safely, on demand, & reproducibly.

Have you ever wondered how to address diversity and increase hit rate in your ML-driven sequential optimization (directed evolution) of enzymes? Small data? No problem, lets select the most informative one! Check out our work featured on front cover with Tobias (@t_vornholt)!

Navigating sequence-activity landscapes with AI and automation - great to have our article featured on the cover of @ACSCentSci

An Artificial Metalloenzyme for Atroposelective Metathesis by Thomas R. Ward, Christof Sparr, Markus Jeschek, and co-workers (@WardGroupBS, @christof_sparr, @JeschekLab) onlinelibrary.wiley.com/doi/10.1002/cc…

🌟#ICLR2024 Spotlight🌟 How to learn a policy in an MDP with non-additive rewards, e.g., experiment design, exploration, etc?🤔 Excited to share our paper: Submodular RL for such challenging tasks; where standard RL fails P: openreview.net/pdf?id=loYSzjS… W: @mutny_ml, MZ, @arkrause

The commentary I wrote is now live at Nature Biotech: nature.com/articles/s4158… (free version rdcu.be/dwSSy).

I am at ICLR 2024. Happy to meet familiar and new faces! If you want to talk about AI, active learning, experiment design and self-driving labs, send me a msg. #AI4Science #ICLR2024

I cannot emphasize the importance of carefully thinking about the particulars of data collection, including some of the ungainly data engineering bits, enough.