F-G Fernandez
Machine Learning Engineer at relaycli pyronear
Paris, Ile-de-France
Summary
🤩
RockstarF-G Fernandez is a Machine Learning Engineer based in Paris with nine years of experience building and hardening deep learning models and tooling. He contributes actively to open-source projects—most notably improving model implementations in docTR (an OCR library) and refining tests across the official PyTorch Vision repo—demonstrating a strong focus on model correctness and testable, production-ready code. His work on torch-cam includes implementing GradCAM variants and shipping visualization tooling, plus automating docs and packaging, showing fluency across research code, developer UX, and release engineering. Comfortable moving between model internals, test automation, and full-stack project setup, he brings pragmatic engineering discipline to ML systems. Colleagues can expect attention to robustness and reproducibility, with a knack for catching tricky parameter-loading and configuration bugs that often hide in deep-learning stacks.
8 years of coding experience
Stackoverflow
66reputation
21kreached
3answers
0questions
Github Skills (28)
unit-testing10
pytorch10
github-ci10
text-recognition10
python10
sphinx10
testing10
ocr10
cicd10
mask-rcnn10
deep-learning10
trainings10
ocra10
computer-vision10
faster-rcnn10
Programming languages (14)
Github contributions (5)
frgfm/torch-cam
Mar 2020 - Jan 2023
Class activation maps for your PyTorch models (CAM, Grad-CAM, Grad-CAM++, Smooth Grad-CAM++, Score-CAM, SS-CAM, IS-CAM, XGrad-CAM, Layer-CAM)
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:8 releases, 29 reviews, 258 commits in 2 years 10 months
Contributions summary:F-G was primarily involved in the implementation of core functionality related to class activation maps (CAMs) within the PyTorch framework. They implemented GradCAM and GradCAM++ extractors and their related visualization utilities. Furthermore, the user contributed significantly to project setup by adding package initialization, documentation using Sphinx, and test configurations, along with the integration of automated workflows for testing, documentation deployment, and PyPI publishing.
pythonscore-camsaliency-mapcamclass-activation-maps
mindee/doctr
Apr 2022 - Jul 2022
docTR (Document Text Recognition) - a seamless, high-performing & accessible library for OCR-related tasks powered by Deep Learning.
Role in this project:ML Engineer
ML EngineerContributions:137 reviews, 21 commits, 15 PRs in 3 months
Contributions summary:F-G primarily contributed to fixing and improving model-related components within the `doctr` library. Their work involved correcting errors in loading pretrained parameters, addressing typos, and refining model configurations for various architectures such as ResNet, MobileNet, and CRNN. They also updated tests and made refactoring changes to improve the robustness of the model implementation, especially in the context of PyTorch.
ocr-recognitiontext-documentoptical-character-recognitiontext-recognitioncomputer-vision
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.