Staff AI Research Engineer, Data Infrastructure

Roboforce Inc.

United States

2 months ago

Role details

Contract type

Permanent contract

Employment type

Full-time (> 32 hours)

Experience level

Expert

Experience required

5 years minimum

Working hours

Regular working hours

Job source

dice.com

Tech stack

Amazon S3 Automated Storage and Retrieval Systems Backup Devices Data Deduplication Extract Transform Load (ETL) Distributed Data Store Python (Programming Language) Machine Learning Management of Software Versions Reinforcement Learning Pytorch Deep Learning

+3 more

Information Technology Machine Learning Operations Data Pipelines

Job description

Design and maintain end-to-end data collection pipelines ingesting multimodal demonstration data from teleoperation devices and UMI hardware, including synchronization, versioning, and distributed storage at scale.
Build annotation tooling and data curation workflows - quality filtering, deduplication, episode scoring, and domain reweighting - to produce high-quality training datasets for robot policy learning.
Develop post-SFT reinforcement learning infrastructure: implement reward scoring on demonstrations, mine and categorize failure patterns, and feed curated failure data back into the retraining loop.
Build evaluation and test infrastructure to log policy rollouts on-robot, capture structured results, and surface actionable diagnostics for the research team.
Collaborate with ML researchers to define data schemas, episode formats, and pipeline interfaces that support rapid iteration on VLA and manipulation policy training.
Architect scalable storage and retrieval systems for heterogeneous robot data (vision, proprioception, action, language) across both cloud and on-prem environments.

Requirements

Bachelor’s or Master’s degree in Computer Science, Robotics, or related field with 5+ years of experience.
Strong proficiency in Python and experience building production-grade data pipelines and ETL systems.
Hands-on experience with large-scale dataset management, including versioning, deduplication, quality filtering, and distributed storage (e.g., S3, GCS, HDF5, WebDataset, Zarr).
Experience building or working with post-training infrastructure - SFT pipelines, reward modeling, or RL training loops (e.g., PPO, DPO, rejection sampling).
Familiarity with deep learning frameworks (PyTorch, JAX) and ML training workflows sufficient to collaborate tightly with research teams.
Requires 5 days/week in-office collaboration with the teams., * Experience with robotics data collection hardware - teleoperation devices, UMI, GELLO, or similar - and the synchronization and preprocessing challenges they introduce.
Familiarity with robot learning pipelines: imitation learning, behavior cloning, or VLA/VLM fine-tuning workflows.
Experience building evaluation or experiment tracking infrastructure (e.g., Weights & Biases, MLflow, custom rollout loggers).
Proven ability to design annotation tooling or human-in-the-loop labeling systems for structured or multimodal data.