Computer Vision Engineer

Lili-o; Paris; Sur place; EUR 70 000 - 110 000; Plein temps

Résumé du poste

Lili-O in Paris is seeking a Computer Vision Engineer to push forward our one-shot learning approach, collaborating with our science team to build vision models and dense descriptors that operate in real time on robot hardware. You will work on affordance learning, dense descriptors, and real-time flow matching, partnering with researchers to turn results into deployable vision modules on real hardware.

You will research and develop object segmentation across unstructured environments and

Qualifications

• PhD or strong research background in computer vision / machine learning appreciated.
• Experience taking vision research from prototype to real-time on-robot deployment is a plus.
• Comfortable working in a purely vision-based paradigm with no reliance on markers or hand-crafted priors.

Responsabilités

• Research and develop vision models for one-shot learning of new objects and tasks.
• Work on affordance learning - teaching the robot where and how it can interact with an object from a single demonstration.
• Develop and refine dense descriptors for robust object/keypoint representation across viewpoints.
• Build object segmentation pipelines that hold up in unstructured, real-world environments.
• Explore flow matching approaches for vision-driven generative/representation models.
• Optimize models for real-time inference on robot hardware.
• Collaborate closely with the research team on research direction, experiments, and turning results into shippable vision modules.

Connaissances

One-shot learning

Affordance learning

Dense descriptors

Real-time inference

Flow matching

Python

Deep learning

Neural network architectures

Formation

PhD or strong research background in CV/ML

Outils

PyTorch

Description du poste

Push forward our one-shot learning approach alongside our science team - build vision models, affordance learning, and dense descriptors that run in real time on robot hardware.

Focus Computer vision robot learning one-shot learning

About Lili-O

Our mission is to empower every person with a robot - whether that’s an elderly person who needs a hand at home or a young professional who wants their time back.

To get there, we’re building the brain that lets robots automate specific long-horizon tasks in unstructured, real-world environments.

That brain combines skill learning, vision, and long-horizon planning - built on a skill-based approach that decomposes complex tasks into reusable primitives.

The Role

As a Computer Vision Engineer, your core mission is to push forward our one-shot learning approach, working side by side with our scientist team.

Responsibilities

• Research and develop vision models for one-shot learning of new objects and tasks
• Work on affordance learning - teaching the robot where and how it can interact with an object from a single demonstration
• Develop and refine dense descriptors for robust object/keypoint representation across viewpoints
• Build object segmentation pipelines that hold up in unstructured, real-world environments
• Explore flow matching approaches for vision-driven generative/representation models
• Optimize models for real-time inference on robot hardware
• Collaborate closely with the research team on research direction, experiments, and turning results into shippable vision modules

What We’re Looking For

Hard Skills

• Affordance learning / affordance knowledge
• Dense descriptors (e.g. for correspondence, keypoint, or pose representation)
• Real-time inference and model optimization
• Flow matching
• Strong Python and deep learning fundamentals (PyTorch or equivalent)
• Ability to design, implement, and train novel neural network architectures from scratch, not just fine-tune existing ones

Background

• PhD or strong research background in computer vision / machine learning appreciated, ideally with publications
• Experience taking vision research from prototype to real-time, on-robot deployment is a plus
• Comfortable working in a purely vision-based paradigm - no reliance on markers, motion capture, or hand-crafted priors

Soft Skills

• Strong research collaboration skills - this role works hand-in-hand with our chief scientist
• Curiosity and rigor, balanced with a drive to get things running on real hardware
• Comfortable iterating quickly on open research questions

Practical Information

• Location: Paris / South Korea
• Contract type: Full-time

Work Culture - What to Expect

This role is for someone who thrives with autonomy and ambiguity, not someone who needs a roadmap handed to them every morning.

• You’ll often work independently. Long stretches of focused, self-directed research and experimentation, with limited day-to-day supervision. We trust you to figure out the path forward.
• Travel is part of the job. Conferences, labs, partner visits - if there’s a relevant event abroad, we expect you to be comfortable hopping on a plane to represent Lili-O and bring ideas back.
• Priorities shift, fast. This is a startup. We will sometimes change direction and de-prioritize work you’ve put real effort into. It’s never personal - it’s how startups survive.
• Comfort with uncertainty. Specs, scope, and even the problem definition will evolve. You’ll be expected to bring structure to that ambiguity rather than wait for it to be resolved for you.

If you get energy from solving open problems, adapting on the fly, and owning your slice of the robot end-to-end - let’s talk.