Tor Fjelde
AI Research Scientist at University of Cambridge
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolTor Fjelde is an AI research scientist and PhD researcher at the University of Cambridge with 11 years of experience building high-performance backend systems and probabilistic ML tools. With a first-class Mathematics degree from Edinburgh, he blends rigorous theoretical skills in math and physics with practical programming and data-analysis expertise applied to variational inference and probabilistic programming. He contributed significant improvements to the widely used Turing.jl library—adding AD-backend support, Flux optimizers, and ELBO/ADVI enhancements—demonstrating a rare combination of research-grade algorithms and production-oriented engineering. Now at Xaira Therapeutics, he applies computational learning methods to real-world problems, drawing on internships and research stints including Microsoft Research Cambridge. Colleagues describe him as equally comfortable deriving math on a whiteboard and shipping robust backend code.
10 years of coding experience
2 years of employment as a software developer
Bachelor's degree, Mathematics, 1st, Bachelor's degree, Mathematics, 1st at The University of EdinburghDoctor of Philosophy - PhD, Doctor of Philosophy - PhD at University of Cambridge
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (9)
machine-learning10
variational-inference10
julialang10
bayesian10
bayesian-inference10
julia10
probabilistic-programming9
mcmc7
artificial-intelligence6
Programming languages (16)
Github contributions (5)
TuringLang/Turing.jl
Jul 2019 - Jan 2023
Bayesian inference with probabilistic programming.
Role in this project:
Back-end Developer / ML Engineer
Back-end Developer / ML EngineerContributions:239 reviews, 107 commits, 136 PRs in 3 years 6 months
Contributions summary:Tor made significant contributions to the `turing.jl` repository, focusing on variational inference and related algorithms. They implemented a rough version of variational inference, modified the directory structure, and adapted the interface. The user added support for different AD-backends, incorporated Flux optimizers, and improved the ELBO implementation. They also improved the ADVI implementation and updated the code to integrate with the new Bijectors interface.
julia-languagebayesian-inferencebayesian-methodsbayesian-neural-networkshamiltonian-monte-carlo
torfjelde/Bijectors.jl
Jul 2019 - Sep 2020
Automatic transformations for constrained random variables.
Contributions:1 PR, 276 pushes, 27 branches in 1 year 2 months
random-variablesconstrainedtransformationsvariables
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.