Software Engineer, Robotics

Deccan AI • San Francisco, CA

8d • Onsite

About The Position

Deccan AI is a model training and evaluation startup located in Mountain View, CA, with a delivery center in Hyderabad. Founded by alumni from IIT Bombay, IIM Ahmedabad, and ex-Google, and backed by Prosus Ventures, the company specializes in building expert-curated datasets and evaluation infrastructure for leading AI labs. Their Physical AI practice focuses on creating the data backbone for embodied AI, including annotation, synthetic data generation, and model evaluation to accelerate robot learning in manipulation and reasoning.

The VLA (Vision-Language-Action) foundation models require robust production software to integrate data pipelines, simulation, training loops, and evaluation into a cohesive system. This role is situated at the intersection of ML research and systems engineering, where the engineer will develop the end-to-end systems that enable VLA training. The position involves direct collaboration with clients from frontier labs and shipping systems critical for training next-generation robot foundation models.

Requirements

• MS or PhD in Computer Science, Robotics, or Machine Learning, or equivalent industry experience.
• 2+ years of experience shipping production systems, beyond prototypes.
• Strong Python and systems skills.
• Comfortable with Linux, Docker, CUDA, and cloud infrastructure (AWS/GCP).
• Working knowledge of robot learning, including imitation learning, behavior cloning, diffusion policies, or reinforcement learning.
• Understanding of how VLA training works, not just the theory.
• Experience with at least one robotics simulator (Isaac Sim, MuJoCo, PyBullet, Gazebo).
• Experience with robotics data formats/middleware (ROS/ROS2, URDF/USD, MCAP, HDF5).
• Experience writing clean, tested, and documented code.
• CI/CD and code review experience.

Nice To Haves

• Hands-on experience with VLA codebases such as RT-2, Octo, OpenVLA, π0, GR00T N1, or LeRobot.
• Experience with NVIDIA Omniverse, Replicator, or Cosmos for synthetic data generation.
• Familiarity with sim-to-real transfer techniques like domain randomization, NeRF, or Gaussian Splatting.
• Experience building ML evaluation frameworks, model benchmarking suites, or RLHF/DPO training pipelines.
• Open-source contributions in robotics or machine learning.
• Published research in manipulation, embodied AI, or VLA systems.

Responsibilities

• Build the VLA training integration layer, including data loaders, format converters, and preprocessing steps to feed curated datasets (real and synthetic) into training frameworks like LeRobot, Octo, OpenVLA, and π0.
• Develop evaluation and benchmarking infrastructure, encompassing sim-based rollouts in Isaac Lab, success-rate tracking, regression detection, and automated reporting for client delivery.
• Own dataset management, including versioning, schema validation, metadata indexing, and format conversion across various formats such as Open X-Embodiment, LeRobot HDF5, RLDS, and client-specific formats.
• Implement sim-to-real transfer tooling, such as domain randomization configurations, Cosmos Transfer integration, and quality validation to ensure synthetic data enhances policy performance.
• Build annotation platform backend systems, including task taxonomy APIs, temporal segmentation, quality scoring, and inter-annotator agreement mechanisms for robotics episode data.
• Integrate with the NVIDIA Isaac ecosystem (Isaac Sim, Isaac Lab, GR00T-Mimic, OSMO) to orchestrate synthetic data generation on cloud GPU infrastructure.

Benefits

• Founding-stage impact: Opportunity to shape the robotics practice at its inception and define how Deccan delivers physical AI data for years to come.
• Work with frontier lab clients: Direct contribution to systems used by teams like Google DeepMind, placing you at the forefront of embodied AI.
• Full-stack ownership: Responsibility for entire systems, from annotation backends and VLA training integration to sim-based evaluation, rather than isolated tasks.

What This Job Offers

• Job Type: Full-time
• Career Level: Mid Level