Paul Balança
Research Team Lead at Graphcore
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolPaul Balança is a research team lead and machine learning engineer with 14 years of experience blending deep theoretical probability and fractal analysis with production ML for autonomy and hardware acceleration. He has driven sensor-fusion and object-detection improvements in autonomous driving at FiveAI, led low-precision LLM and DFT-on-IPU research at Graphcore, and maintains core JAX/IPU tooling used for high-performance model execution. Comfortable from mathematical proofs to production pipelines, he has published on multifractal structures of stochastic processes while shipping real-time perception stacks and data pipelines. A hands-on engineer and former startup CTO, he also contributes practical open-source work such as an SSD TensorFlow implementation and IPU-focused JAX libraries, reflecting a rare mix of rigorous math and systems-level ML craftsmanship.
14 years of coding experience
8 years of employment as a software developer
Master of Engineering (Grande Ecole) Applied Mathematics, Master of Engineering (Grande Ecole) Applied Mathematics at CentraleSupélecMaster 2 Mathematical Statistics and Probability, Master 2 Mathematical Statistics and Probability at Paris-Sud University (Paris XI)Udacity
French, English
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (5)
tensorflow10
python10
machine-learning9
computer-vision9
deep-learning8
Programming languages (4)
Github contributions (5)
balancap/SSD-Tensorflow
Jan 2017 - Apr 2017
Single Shot MultiBox Detector in TensorFlow
Role in this project:
Back-end Developer
Back-end DeveloperContributions:122 commits, 3 PRs, 87 pushes in 2 months
Contributions summary:Paul's primary contributions focused on the foundational structure of the SSD TensorFlow project. Their commits established the basic structure for datasets, networks, and pre-processing components. They implemented essential parts of the network, including the VGG-based SSD model, and configured the loss functions. Furthermore, they worked on the dataset conversion and annotation, and implemented some basic tests using Jupyter Notebooks.
single-shot-multibox-detectormultibox-detectordetectorcaffetensorflow
balancap/models
Sep 2016 - Nov 2016
Contributions:34 commits, 24 pushes, 2 branches in 1 month
tensorflow
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.