Research Scientist / Research Engineer, Models

The Role

At Config, we're building the physical intelligence layer for the next generation of robots.

Our mission is ambitious: to bring robot foundation models out of research labs and into the real world, where robots can reliably perform useful work for customers. We are looking for exceptional researchers and research engineers to work on the core model and learning systems behind Config's robot foundation models.

You will work on high-impact problems at the intersection of robot learning, multimodal and multi-sensor foundation models, robot control, and real-world robotics deployment. Your work may span policy model architecture, multi-sensor and multimodal fusion, training methodology optimization, model inference optimization, multi-embodiment data conversion, action representation, controller-policy interfaces, and real-world robustness.

This is a hands-on research and engineering role for people who want their work to matter beyond benchmarks. You will not only explore ideas in isolation — you will help build models and systems that are tested on real robots, real tasks, and real-world variations.

If you are excited about pushing robot foundation models from research toward production-grade physical intelligence, we'd love to meet you.

** This JD is designed to cover two role tracks across multiple levels.

Level: We are open to candidates from interns to senior, staff, and senior staff professionals. Level, scope, ownership, and compensation will be calibrated based on experience.

Role Track: We are hiring for both Research Scientist and Research Engineer tracks. Research Scientists focus more on developing and validating new ideas, hypotheses, model architectures, and learning methods. Research Engineers focus more on building, scaling, and validating the pipelines that turn those methods into real working systems.

When applying, please specify the role track and level that best reflects your experience and interests. We're happy to calibrate scope and leveling throughout the interview process.

Areas You May Contribute To

The examples below reflect some of the high-priority research and engineering directions at Config today. They are not meant to be exhaustive — we are excited to shape the scope around exceptional candidates who can identify and drive other high-impact problems in robot foundation models.

• Robot Foundation Model Architecture — Design policy architectures that can fuse multimodal signals — including written and spoken language, multi-sensor observations, robot states, and action representations — to generate robust robot actions with reasoning capability.
• Model Training Methodology — Improve training recipes for robot foundation models, including imitation learning, reinforcement learning, data mixing, loss design, sampling strategy, and curriculum learning across diverse tasks and embodiments.
• Model Inference & Deployment Optimization — Optimize real-time policy inference so models can run on physical robots with low latency, stable actions, and reliable closed-loop behavior.
• Multi-Embodiment Data & Action Conversion — Develop methods to convert human demonstration data and robot trajectories into action representations that can transfer across different robot embodiments.
• Robot Control & Policy Interface — Improve the interface between learned policies and robot controllers so model outputs translate into smooth, safe, and reliable robot motion.
• Evaluation & Robustness — Build scalable evaluation frameworks that measure robot foundation model performance under real-world and simulation-based variation.

What We're Looking For

We are open to candidates across levels, from exceptional research interns to experienced senior+ candidates. The level, scope, ownership, and compensation will be calibrated based on experience.

We are also open to two role tracks: Research Scientist and Research Engineer. We do not expect every candidate to be equally strong in both tracks, but we do expect strong technical depth, hands-on execution, and a deep interest in building robot foundation models that work in the real world.

Shared Requirements

• Strong motivation to work on foundation models, physical AI, and real-world robot learning
• Strong programming ability, especially in Python
• Experience with deep learning frameworks such as PyTorch or JAX
• Solid understanding of machine learning fundamentals
• Ability to read research papers, implement ideas, run experiments, and communicate results clearly
• Interest in robot learning, imitation learning, reinforcement learning, multimodal learning, computer vision, control, or robotics systems
• High ownership, curiosity, and urgency in solving difficult technical problems
• Excellent communication and collaboration skills in Korean and written English

For senior-level candidates, we typically look for 5+ years of relevant research or engineering experience, or equivalent evidence of exceptional technical ownership. For staff and senior staff-level candidates, we look for a stronger track record of independently defining technical direction, driving ambiguous projects, mentoring others, and delivering high-impact systems or research outcomes.

Research Scientist Track

We are especially interested in candidates who can define and drive research directions in robot foundation models. Strong signals include:

• Experience formulating research hypotheses, designing controlled experiments, and analyzing model behavior
• Strong understanding of model architectures, training methodologies, action representations, policy learning, or multimodal learning
• Ability to identify high-leverage research problems from real robot failures, data limitations, or model performance gaps
• Experience developing new methods for robot policy learning, multi-sensor fusion, multi-embodiment learning, inference-time adaptation, or real-world generalization
• Research track record through publications, strong preprints, open-source projects, competition results, or equivalent technical work

For senior+ Research Scientist candidates, strong signals include independently defining research agendas, developing new methods, designing rigorous experiments, and turning research insights into measurable improvements in robot performance.

Research Engineer Track

We are especially interested in candidates who can turn research ideas into reliable systems. Strong signals include:

• Experience building training/inference pipelines, data pipelines, model evaluation systems, or large-scale experimentation infrastructure
• Strong implementation skills for debugging models, scaling experiments, optimizing training, and improving inference performance
• Experience with distributed training, GPU optimization, dataset tooling, model serving, latency optimization, or real-time inference
• Ability to build robust systems that connect model outputs to robot execution, controller interfaces, data logging, and evaluation workflows
• Strong engineering judgment in making research code reliable, scalable, and usable by the broader team

For senior+ Research Engineer candidates, strong signals include owning major training, inference, data, evaluation, or deployment systems; scaling research ideas into reliable infrastructure; and building systems that make the broader team faster and more effective.

Above all else, we care about intellectual intensity, first-principles thinking, hands-on execution, and the desire to solve problems that bring AI into the physical world.

Bonus Points

We'd be especially excited if you have:

• Experience with Vision-Language-Action models, world models, robot foundation models, imitation learning, or reinforcement learning
• Experience training or fine-tuning large-scale deep learning models
• Experience with multimodal learning, multi-view vision, video models, sensor fusion, or temporal modeling
• Experience with multi-embodiment learning, embodiment-conditioned models, or cross-embodiment data conversion
• Experience with distributed training, large-scale datasets, GPU optimization, or model inference optimization
• Experience with robotics simulation tools such as MuJoCo, Isaac Sim, Gazebo, or similar platforms
• Experience with ROS or ROS2
• Experience with real robot experiments, teleoperation systems, or data collection pipelines
• Strong research track record, open-source projects, competition experience, or exceptional engineering projects
• Prior experience in fast-moving startup or research-driven engineering environments