Ivar Flakstad
Machine Learning Engineer at Hugging Face
Oslo, Norway
Summary
🤩
Rockstar🎓
Top SchoolIvar Flakstad is a Machine Learning Engineer based in Oslo with 12 years of software engineering experience spanning backend systems, ML infrastructure, and test automation. Currently at Hugging Face, he brings practical expertise from roles at Curieo Ai and Oda where he blended software engineering and ML responsibilities to ship production-ready systems. He is an active open-source contributor to high-profile projects like Hugging Face Transformers—focusing on test robustness across many models—and has implemented performant random number generation and benchmarking for the Rust-based Candle framework. Comfortable across languages and runtimes, he has hands-on experience integrating native kernels (Metal) with higher-level ML tooling, revealing a deep interest in performance-sensitive ML primitives. With a Computer Science background from NTNU and a track record in consulting and public-sector mapping services, he pairs rigorous engineering discipline with pragmatic product focus. Known for improving reliability at the intersection of ML models and infrastructure, he often surfaces subtle correctness issues that prevent regressions in production.
12 years of coding experience
7 years of employment as a software developer
Electronics and Computer Technology, Electronics and Computer Technology at University of OsloNorwegian University of Science and Technology
English, Norwegian
Github Skills (15)
testing10
randomization10
gpu-programming10
pytorch10
machine-learning10
rust10
model-testing10
python9
linear-algebra9
benchmarking9
nlp9
transformer8
bert6
tensorflow3
jax3
Programming languages (4)
Github contributions (5)
huggingface/candle
Aug 2023 - Mar 2025
Minimalist ML framework for Rust
Role in this project:
ML Engineer
ML EngineerContributions:37 reviews, 20 PRs, 69 pushes in 1 year 7 months
Contributions summary:Ivar implemented a hybrid Tausworthe + LCG pseudorandom number generator in metal, designed for use in the "candle" ML framework. They subsequently added a Gaussian normal distribution based on the Box-Muller transform, enhancing the framework's random number generation capabilities. This involved modifications to the random.metal kernel and integration with existing Rust code, enabling uniform and normal random number generation within the Metal backend of the framework. Further contributions include the addition of benchmarks for evaluating the performance of the random number generators, which helps to assess the impact of these changes.
huggingface/transformers
Dec 2024 - Apr 2025
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:QA Engineer / Test Automation Engineer
QA Engineer / Test Automation EngineerContributions:12 reviews, 39 PRs, 118 pushes in 3 months
Contributions summary:Ivar primarily contributed to improving the quality and robustness of the `transformers` library by adding and modifying tests. Their commits focused on addressing issues in model integration tests, including fixing incorrect or missing tolerance values and correcting test logic. They made adjustments to tests for several models, including Autoformer, Aria, WavLM, TimeSeriesTransformer, Informer, Git, and Bamba, indicating a broad understanding of the library's model offerings.
pythonbertspeech-recognitionstate-of-the-artflax
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.