Torsten Scholak
Member Of Technical Staff at Cohere
Montreal, Quebec, Canada
Summary
🤩
Rockstar🎓
Top SchoolTorsten Scholak is a research-driven software engineer with 11 years of experience building foundation-models and conversational assistants, currently splitting time between leading the Foundation Models Lab at ServiceNow Research and as Member of Technical Staff at Cohere. He blends deep academic training—a PhD in theoretical and mathematical physics—with practical ML and systems work, from GPU-accelerated scientific computing to production research on large language models. Torsten’s career spans applied research and engineering roles at Element AI and ServiceNow, contributing both conceptual advances and production-ready systems that bridge research and product. He is an active open-source contributor in niche areas like HaskTorch, improving typed neural network components and doctests—an indicator of attention to correctness in tooling not obvious from his job titles. Comfortable with low-level performance optimization and high-level model design, he consistently turns complex mathematical ideas into scalable software. Based in Montreal, he pairs scientific rigor with hands-on engineering to drive reliable conversational AI.
11 years of coding experience
20 years of employment as a software developer
Doctor of Philosophy (Ph.D.), Theoretical and Mathematical Physics, Doctor of Philosophy (Ph.D.), Theoretical and Mathematical Physics at The University of FreiburgDiplom (German equivalent of M.S. degree), Theoretical and Mathematical Physics, Diplom (German equivalent of M.S. degree), Theoretical and Mathematical Physics at University of Bayreuth
German, English
Github Skills (4)
haskell10
pytorch9
automatic-differentiation9
testing7
Programming languages (21)
Github contributions (5)
hasktorch/hasktorch
Aug 2019 - Oct 2021
Tensors and neural networks in Haskell
Role in this project:
Back-end Developer
Back-end DeveloperContributions:120 reviews, 822 commits, 167 PRs in 2 years 2 months
Contributions summary:Torsten's contributions focused on removing labels from a `MultiheadAttentionSpec` in the `Torch/Typed/NN/Transformer.hs` file, and fixing doctests in various `Torch/Typed/Functional.hs` and `Torch/Typed/NN/Recurrent/LSTM.hs` files. The user also implemented a 'Generic HasForward idea' in relation to a `Torch/NN.hs` file and addressed a number of merge conflicts. These changes indicate the user worked on improving the code base.
haskellneural-networksmachine-learningneural-networktensors
hasktorch/hasktorch-skeleton
Jun 2020 - Apr 2021
Nix Skeleton Hasktorch project made for easy cloning and forking
Contributions:3 reviews, 37 commits, 3 PRs in 9 months
forkingcloningnixosskeleton
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.