Research Engineer — Vision & Perception
Bleu Robotics
Bleu Robotics is an AI company developing universal AI software for the deployment of advanced robots , humanoids, mobile manipulators, and industrial arms, in real factories. The convergence of humanoid robotics and modern AI is one of the most significant technical opportunities of the decade, and most of the work today still happens in research settings. We are building the software layer required to bring it into operational use.
We have a strong ambition to become the next Microsoft to run on every industrial humanoid. We have closed our pre-seed round with Top-tier fund to deploy our robot in Tier 1 manufacturing facilities.
We are looking for deeply passionate scientists who can help us push the frontier and build a major lab in our physical AI.
Our vision
To enable a robot to learn a new industrial task from a few demonstrations and operate autonomously, with under one hour between the start of the demonstration and the start of deployment.
This objective shapes our roadmap and every technical decision we make.
The role
We are hiring Research Scientists in Vision & Perception to design the perception stack that makes our robots understand their environment well enough to act in it reliably. Your work will define how our robots see the world, what they recognize, how robustly, and from how little data.
Industrial environments are visually complex, varied, and unforgiving: changing lighting, cluttered scenes, novel objects, and a hard requirement to work in production without supervision. The perception layer has to deliver the spatial and semantic understanding required for learned manipulation policies to succeed, across the many tasks, objects, and sites we will deploy on. Hitting our one-hour demonstration-to-deployment target depends directly on perception methods that work robustly out of the box, with minimal site-specific tuning.
This is a research role focused on shipping algorithms into a real product. We are an early-stage company and do not publish at this stage. You will work in close collaboration with the Robot Learning and Integration teams to ensure the methods you develop run on real robots, in real factories.
Responsibilities
- Design, prototype, and evaluate perception algorithms for robotic manipulation — object recognition, segmentation, pose estimation, scene understanding, and related areas
- Advance our internal state of the art in perception for embodied agents, with a focus on robustness and data efficiency
- Run rigorous experiments on real robots, and contribute to the data, evaluation, and infrastructure needed to do so (we will support you in ramping up on robotics if your background is primarily in vision)
- Translate research prototypes into perception components robust enough to ship as part of our product
- Collaborate with the Robot Learning and Integration teams to move research from prototype to production
- Help shape the long-term research agenda of the company
The engineering team is organized around three focus areas with multiple contributors each: Integration, Vision & Perception (this role), and Robot Learning. Early hires will help shape the research agenda, engineering practices, hiring, and culture.
Required Qualifications
- PhD or MSc in computer vision, robotics, machine learning, or a closely related field, from a leading research lab or strong experience in research environments
- Track record of published research at top venues such as CVPR, ICCV, ECCV, NeurIPS, ICML, ICRA, IROS, CoRL, RSS, or equivalent
- Demonstrated research expertise in computer vision or machine perception — object recognition, segmentation, pose estimation, scene understanding, 3D perception, or representation learning for vision
- Strong Python skills and fluency with PyTorch
- Hands-on experience training, evaluating, and debugging modern vision systems
- Professional proficiency in English (internal working language)
The strongest candidates will combine deep research expertise with experience across several of the following:
- 3D vision — depth estimation, geometric reasoning, point cloud processing, or 3D scene reconstruction
- Perception for robotic manipulation, including grasping and interaction
- Experience working with real robots (any platform — humanoids, arms, mobile bases)
- Postdoctoral experience, or research positions in industry labs or top academic groups
- A history of translating research into deployed systems
- A clear, ambitious research mission with measurable real-world impact
- A diverse fleet of advanced robots, including humanoids, on which to run your research
- Research that ships: your algorithms run on real machines in real factories
- Tight collaboration with engineering teams that take your work to production
- Significant ownership of the research agenda as an early member of the team
- A focused, technically strong team and a serious engineering culture
- English-speaking working environment, no French required
- Fair compensation + equity