Paul Balança
Research Team Lead @Graphcore
Signup · Get unlimited contacts
WORK HISTORY
Research Team Lead @Graphcore
London, GB
Research Team Lead in Graphcore.Main research projects- Low floating point precision for LLM inference & training (1 conference paper)- Molecular dataset generation using density functional theory (DFT) on IPUs (2 conference papers)- Low level functional Python programming on IPUs;Software contribution- Main developer & maintainer of JAX on Mk2 IPUs (https://github.com/graphcore-research/jax-experimental)- TessellateIPU: exposing low level IPU programming in Python (https://github.com/graphcore-research/tessellate-ipu)- JAX Scalify: end-to-end scale arithmetic for low precision training (https://github.com/graphcore-research/jax-scalify)
EDUCATION
Paris-Sud University (Paris XI)
Master 2, Mathematical Statistics and Probability
CentraleSupélec
Master of Engineering (Grande Ecole), Applied Mathematics
SKILLS
ABOUT PAUL BALANÇA
Staff Machine learning Research Engineer at FiveAI - Working on Multi-sensor Fusion and Object Detection. I have interest and deep experience in real world application of state of the art Machine Learning, as well as theoretical foundations.I am originally a researcher in Probability Theory, loving topics connected to (multi)fractal geometry and functional analysis (funny functional spaces and wavelets descriptions).During my academic experience, I have worked on several modern probability topics, including Lévy processes, continuous stable trees and stochastic partial differential equations (SPDEs), and solve a few open questions mostly related to the multifractal structure of the former.I have been interested the last couple of years in more applied mathematics related to machine learning and image processing, and aim to push further the connections between these topics and my theoretical research.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.