Björn Mattsson
I have a BSc and MSc in physics with courses in computational biology, worked in machine learning for a decade, and now provide consulting services in the field of computational chemistry. I am technology agnostic and believe that both modern ML methods and physics-based methods are valuable; the hard part is finding the right combination of tools for the problem at hand. I also believe that rigorous benchmarking is fundamental to make progress, both in individual projects and as a field.
Biography
Computational chemistry consultant
I started Enlace Bio in 2024 to provide computational chemistry services to biotechs and startups.
Public work:
-
Co-published a new benchmark for protein-ligand affinity prediction with Pat Walters (OpenADMET): Identifying and Addressing Systematic Data Leakage in Protein-Ligand Affinity Benchmarks, Novelty-Tiered Affinity Benchmark (NTAB).
-
4/100 placement in the OpenADMET blind PXR induction structure prediction competition, by combining co-folding with physics-based pose selection: competition closing announcement, my method.
-
14/350 placement in the OpenADMET blind PXR induction activity prediction competition, by combining ML, co-folding and semi-empirical quantum mechanics methods: competition closing announcement, my method.
Professional experience in machine learning
I have spent a decade building ML products. At Tractable I as Head of Estimating AI helped build an AI-based computer vision product from early prototype to a globally deployed software platform, resulting in nine co-invented US patents. That experience gave me a practical understanding of what it takes to build ML systems for real-world usage.
Background in physics
My BSc thesis in computational quantum physics was published (>200 citations): Uncertainty analysis and order-by-order optimization of chiral nuclear interactions.
Want to learn more?
Explore some of my code on Github.
Look at my Google Scholar profile.
Connect with me on LinkedIn.
Or get in touch below.