R&D Scientist - 3D Computer Vision

Our missionA robot that doesn't know where it is can't reliably navigate, plan, or act. Localization is the foundation for everything else, yet it remains an open problem: ports and tunnels with no usable GNSS, warehouse aisles that look identical in every direction, construction sites whose geometry changes daily, scenes saturated with moving objects that corrupt the very map you're building from them.At Exwayz, we build the LiDAR perception stack that makes reliable autonomy possible: real-time SLAM and localization at sensor rate, centimeter-level accuracy, robustness to geometric degeneracy and dynamic scenes, and sensor-agnostic performance across LiDAR brands and scan patterns. On top of that foundation, we're building the perception layer that turns raw point clouds into something a robot can act on: detection, segmentation, mapping, and change detection.What we care about is generality: methods that remain sensor-agnostic and keep working on real-world data beyond the distribution of public datasets. That's the bar we set for our own work.We're a team of 8, already in production with clients across Europe and the US.Your roleYou'll work across the perception stack: semantic and panoptic segmentation, 3D object detection, tracking, and the problems that come with pushing all of it into production. The scope is broad by design: we're a small team building a full perception layer, so alongside your own areas of depth, you'll need to understand how the pieces fit together and help make the stack better as a whole, not just your corner of it.Concretely: keeping up with the literature, deciding what's worth trying, training and evaluating it on our data, and taking what works all the way to code that runs in production on a robot, working closely with the rest of the team at each step.This is not a pure research role and not a pure engineering role. We need someone who can read a research paper, tell the difference between a real improvement and a benchmark artifact, and turn the ones that hold up into a working prototype. The ideas matter, but so does the fact that they ship.ResponsibilitiesDesign, train and evaluate deep learning models across the perception stack, on our internal and public LiDAR datasetsOwn perception components end to end, from literature review to deployed codeBuild and maintain the training, evaluation and data pipelines these models depend onBenchmark rigorously: ablations, failure-case analysis, honest baselines, and metrics that reflect what the client actually experiencesCommunicate results and trade-offs clearly to the rest of the team, and to clients when neededOptimize models for real-time inference under embedded compute constraintsDebug perception failures on real client data, in conditions no dataset anticipatedContribute to the technical direction of the perception roadmapMentor interns and share what you learn with the rest of the teamContribute to research paper writing if desiredStackPython, PyTorch, C++, ROS, CUDA, Git, Linux.Candidate requirementsRequired3+ years of industry experience in 3D perception, computer vision or a closely related field, or a PhD in one of these areasStrong deep learning fundamentals, and a track record of models you took from idea to something that worked on real dataHands-on experience with LiDAR point cloudsProduction-quality Python and PyTorch, and enough C++ to work in a real-time codebaseAble to read, critique and reproduce state-of-the-art papers, and to judge which ones are worth the effortAutonomy: you can be handed an open-ended problem and come back with a defensible answerWorking French and EnglishNice to haveExperience across several perception tasks rather than a single specialtyTemporal consistency and tracking in 3DSLAM, registration or sensor fusionModel optimization and deployment on embedded targetsPublications at top-tier vision or robotics venuesExperience working close to clients and their dataConditionsPermanent contract (CDI)Based at our Saint-Lazare office in Paris, mostly on-site with some remote flexibilityHealth insurance fully covered by the companyMeal vouchers50% of your Navigo passEligible for stock options (BSPCE) based on performanceCompute resources sized for real training runs, including HPC accessProcess30' intro call with the Head of AITechnical case study with the Head of AI and the CTOLunch and team meeting at our Saint-Lazare office, and dedicated time with the CEOKeywordsDeep Learning · Computer Vision · Machine Learning · Artificial Intelligence · Neural Networks · Robotics · Autonomous Vehicles · 3D Vision · Multi-Object Tracking · Scene Understanding