The posting
Company
Adoc Talent Management is supporting the recruitment of a Research Engineer – Physical AI for a San Francisco-based company, part of Y Combinator’s 2026 cohort, whose work focuses on collecting human and robotic data to train robots.
The company specializes in collecting first-person, real-world data at a scale and speed that research laboratories cannot readily achieve independently. To ensure traceability and quality, each contributor’s identity is verified through KYC procedures, and every contribution undergoes several quality-control checks before reaching clients.
Looking ahead, the company aims to play a leading role in deploying robots at scale and making the data needed to advance Physical AI more accessible.
To support this ambition, the company is establishing Hub Labs, a research laboratory in Toulouse, France. With access to a large and varied body of data collected directly from contributors, the laboratory will conduct research to support the future deployment of the company’s robot fleet.
The team develops the infrastructure needed to turn research ideas into results quickly and at scale, transforming those results into data that can be used by teams training future AI models.
Position
As Research Engineer – Physical AI, you will work at the interface between scientific research and engineering. Your role will be to turn research ideas into working implementations, reproduce published methods, run experiments and evaluate results. Working closely with the laboratory's researchers, you will help establish whether an approach works in practice, understand its limitations and generate reliable evidence to guide further investigation.
A central part of your work will involve implementing methods described in scientific papers. You will read and interpret publications, translate their methods into code, reproduce results on relevant datasets and investigate discrepancies between expected and observed outcomes. This will require more than implementing algorithms, you will need to understand the experimental setup, identify potential sources of error and assess what the results actually demonstrate.
You will also design and run experiments in collaboration with researchers. This includes training and evaluating models, investigating technical issues that affect experimental reliability, and improving the speed and robustness of experimental workflows. Depending on the work undertaken, you may operate multi-GPU and multi-node training jobs, work with large models and contribute to the laboratory's training stack, including experiment tracking, evaluation tools and computing-resource management.
Research data and infrastructure will be another important part of the role. You will work with data originating from real-world, multi-camera recordings and help prepare clean, versioned datasets for experimentation. Where relevant, you may also work with simulation-generated data. The objective is to provide researchers with reliable data, tools and experimental workflows that enable them to test ideas efficiently and build on previous results.
The position also connects research with physical systems. Depending on the projects, this may involve working with robotic platforms, teleoperation facilities, robotic cells, simulation environments, sensors or calibration systems. Relevant technical specializations may include distributed training and inference optimization, robotics software, simulation, computer vision, video-based learning or large-scale 3D data. You are not expected to be an expert in every area; the important point is to bring strong engineering skills and depth in at least one relevant field.
Beyond individual experiments, you will contribute to making the laboratory's research reproducible and useful. This may include maintaining reproducible code, datasets and models, documenting experimental methods and results, contributing to scientific publications and making research outputs available through open-source channels where appropriate. You may also help translate successful research into practical capabilities for the wider company, for example by integrating a model into an existing data workflow.
This is a research engineering role in which scientific implementation and experimental rigor are central. It is not primarily a conventional production machine learning position focused on maintaining established pipelines or deploying existing models. You will work closely with researchers to make ideas testable, investigate difficult technical problems and establish what can be learned from experimental results.
Profile
You have a strong technical background in computer science, machine learning, robotics or a related engineering or scientific discipline. You combine solid programming skills with the ability to understand scientific methods and translate them into practical implementations. An engineering degree or master's degree is expected; a PhD is an advantage but is not required.
You have strong Python skills and practical experience with PyTorch. You can read scientific papers, implement the methods they describe and work through the challenges of reproducing results. You understand that a successful implementation requires more than code that runs, you can investigate unexpected outcomes, assess experimental limitations and interpret results critically.
You have experience developing, training or evaluating machine learning models and running experiments in a research or technically demanding engineering environment. Familiarity with GPU-based computing is important, while experience with large models, distributed training, multi-GPU systems or training optimization will be particularly valuable. You are comfortable investigating issues involving data loading, training stability, mixed precision, checkpoints, evaluation or reproducibility.
Your technical specialization may lie in one or more areas relevant to Physical AI. This could include distributed training and inference optimization, robotics software, simulation, computer vision, video-based learning or large-scale 3D data. Experience with tools such as MuJoCo, Isaac or ROS may be useful, depending on your background and the projects undertaken. Experience with sensors, calibration, robotic hardware or embedded deployment is also relevant.
Experience with JAX, C++, or technologies such as ONNX and TensorRT can be an additional advantage. Contributions to scientific publications, open-source projects, research codebases, models or datasets are also valuable. These are not all prerequisites: what matters most is the quality of your technical work, your ability to implement research methods and the depth of your experience in at least one relevant area.
You are rigorous, curious and comfortable working on problems for which the answer is not yet known. You can take ownership of technical investigations, collaborate closely with researchers and communicate your findings clearly. You are motivated by testing ideas, understanding why an approach succeeds or fails, and contributing to research through practical engineering rather than remaining at the level of theoretical proposals.
Professional fluency in English is required for collaboration within an international team. French is an advantage but is not essential.
Career perspective
This is an opportunity to join a research laboratory at an early stage and contribute directly to the technical foundations of its work in Physical AI. You will collaborate closely with researchers, implement scientific methods and help develop the experimental workflows needed to investigate new approaches in machine learning and robotics.
The role offers the opportunity to work with real-world data, research-oriented computing infrastructure and robotic research equipment, while contributing to reproducible experiments, scientific publications and practical technical capabilities. As the laboratory's research program develops, you will be able to contribute your engineering expertise to a range of research questions and technical challenges. Candidates available to start immediately will be particularly appreciated.
The position is based in Toulouse, France, within an international team working across research, machine learning and robotics.
If you are motivated by implementing challenging scientific methods, running rigorous experiments and helping advance AI systems that can understand and interact with the physical world, we invite you to apply to Adoc Talent Management with your CV and salary expectations.
