Founding Research Engineer

present of AI; New York (NY); On-site; USD 120,000 - 180,000; Full time

Job summary

Mbodi AI is seeking a researcher to develop embodied AI for robots, enabling language understanding, learning from demonstrations, and reliable physical-world action. This role emphasizes both research and practical system building, leveraging PyTorch and Python to connect model research with hardware deployment.

You'll work on transforming innovative ideas into working robotic systems, focusing on latency-aware deployment and end-to-end system integration with state-of-the-art models.

Qualifications

• Experience with modern ML, transformers, and foundation models.
• Strong Python and PyTorch skills for end-to-end system development.
• Ability to translate research ideas into working robotic systems.

Responsibilities

• Build AI systems for language understanding in robots and learn-from-demonstrations.
• Deploy models on latency-sensitive hardware and validate in the real world.
• Translate research ideas into practical robotic applications.

Skills

Python

PyTorch

Transformers

Foundation models

End-to-end systems

Job description

This role asks for someone who can build AI systems that let robots understand language, learn from demonstrations, and act reliably in the physical world. A strong candidate brings hands‑on experience with modern machine learning, especially transformers and foundation models, plus the ability to take research ideas from experiment to working robot. The job is research heavy but also very practical, so comfort with Python, PyTorch, and end to end system building matters.

How it fits the AI era

AI is central here: the work involves using and adapting large models for reasoning, perception, and control, so much of the job is about integrating and improving AI rather than hand coding every rule. What stays human is the experimental judgment, the systems thinking to connect models to real hardware, and the ability to decide which research directions are worth pursuing. The role also requires physical world validation, which keeps the work grounded beyond pure software.

How it builds AI experience

Someone in this role would gain direct experience with embodied AI, vision language action models, and agentic systems that plan and recover from errors. They would learn how to fine tune foundation models for robotics and deploy them on latency sensitive hardware. Building this experience could come from projects that combine language models with robot learning, imitation learning, or simulation to real world transfer.

Worth showing

• PyTorch and Python
• Robot learning
• Agentic AI systems

This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.

Mbodi AI builds an embodied artificial intelligence platform that enables operators to train industrial robotic manipulators using natural language and physical demonstrations.