I’m a software engineer with a background in video game development. I now have 10 years of programming experience, with skills ranging from low level optimisation to software architecture. I’m doing research in Deep Learning, with a special interest for its application in creatives industries.
At Dimension Studio, I was involved in the creation of volumetric video from multiview video streams. My focus was initially on acquiring materials through differential rendering for relighting. Later, I worked on neural radiance fields and optimized the architecture to produce dynamic scenes instead of just static ones.
At Mirriad, I was responsible for developing and integrating PyTorch deep learning models for automatic color grading and keypoint detection in the company's compositing software. I also worked on a project to automatically embed advertising into sports broadcasts, utilizing player and ball tracking that ran live on video streams.
As a member of Samsung's Extended Reality team, I have been involved in various projects utilizing AR/VR technologies. In 2019, I developed a meta-learning based approach for generating textures for textile generation, particularly for clothing virtual avatars. In 2020, I played a key role in the development of a pipeline for creating virtual avatars from capture rig inputs. My responsibilities included working on the 3D projection module, the segmentation module, and hair reconstruction. My work on this project was recognized with two awards for Software Architecture and Code Review Processes. In 2021, I successfully ported Openpose to mobile using PyTorch for Android, with support for multi-person face and body keypoint estimation.
I was responsible for designing, implementing and optimizing the machine learning framework for the Fetch virtual machine. I also re-implemented the Word2Vec algorithm in C++ as a compute graph to build a semantic search engine for the Fetch Open Economic Framework. This work was open sourced under the Apache license and a demonstration of the Word2Vec implementation is available at https://github.com/pierre-wilmot/Word2Vec-CPP
As a software developer working on the EUREX exchange, I was responsible for developing and refactoring the codebase. I was also the lead developer for the implementation of the new TRF Baskets feature. My main focus was on decoupling the application logic from the messaging and database components in order to improve unit testing and enable the application to run independently from the rest of the software ecosystem. This resulted in a more robust, easier to understand codebase, and a faster development process.
As a researcher, I focused primarily on parametric neural texture synthesis and style transfer, achieving state-of-the-art results in the field. I published a paper titled “Stable and Controllable Neural Texture Synthesis and Style Transfer Using Histogram Losses”. Additionally, I worked on neural texture weathering, texture segmentation, and single image super resolution. I also developed a neural network runtime from scratch using C++ and OpenCL, which was integrated into the company's product.
At Ubisoft, I was responsible for establishing the mobile build system to compile and run unit tests on both iOS and Android devices. Additionally, as a member of the Ubiservices team, I contributed to the development of a cross-platform library that delivers online features for all Ubisoft games.
At BlackMagic Design, I played a key role in establishing the build system using Python. This system allowed for convenient, one-click builds through a web user interface. The process involved checking out the codebase from a Git repository, compiling it for multiple platforms, performing thorough testing, and packaging the result into a ready-to-ship package.
At Good Game Production, I was responsible for porting the in-house game creation framework from Cocos2d (Objective-C) to Cocos2d-x (C++). This involved ensuring seamless compatibility between the two platforms and streamlining the game development process. With the updated framework, game designers were able to quickly and efficiently create and manage game scenes using a drag-and-drop editor. The scene files could then be easily exported for use in the game engine, enabling faster content creation.
At OutOfTheBit, I was part of the mobile game development team using an in-house cross-platform framework in C++. I played a key role in the creation of three games - Four in a Row, Dots & Boxes, and Sound Ride - within a year and they were published on various platforms, including iOS, Android, and Windows Phone. Each project was developed by a team of one developer and one designer. I am proud to mention that Four in a Row and Dots & Boxes reached the top 1 and top 2 spots in the Apple App Store rankings, respectively.