About this role
Your mission & challengesYou design, build, and operate the end-to-end data infrastructure that turns raw multi-modal robot sensor recordings into ML-ready datasets. Working at the intersection of robotics, edge computing, and cloud data engineering, you own the full pipeline, from physical data collection environments to training-ready formats for robot learning.
Data format & pipeline design: Define and evolve internal data formats for multi-modal sensor streams (video, depth, audio, point clouds, joint states, tactile) with efficient compression.
Edge-to-cloud ingestion & transformation: Build services that orchestrate on-premise recording, validate data quality, transfer large volumes to the cloud, and convert raw data into structured formats for ML consumption.
Infrastructure automation: Provision and manage on-premise compute via infrastructure-as-code, and handle containerized deployments and rolling updates across distributed sites.
Platform services & tooling: Develop backend services (TypeScript, Python), shared libraries, CLI tools, and developer documentation.
Multi-agent annotation: Orchestrate vision-language agents that auto-annotate multi-modal episodes at scale, with human-in-the-loop review to keep labels training-ready.
Cross-functional collaboration: Work closely with hardware, robotics, and AI/data science teams to onboard new sensor modalities and track data throughput KPIs.
What you bring:Backend engineering: Strong Python and TypeScript skills in production backend systems.
Cloud: Hands-on cloud experience, AWS preferred.
Robotics middleware: Experience with ROS 2/DDS or similar robotics middleware.
Media codecs: Familiarity with video/audio codecs and compression (FFmpeg, PyAV, H.264/AV1).
Infrastructure-as-code: Ansible, Docker, and Terraform for fleet management.
DevOps mindset: You care about CI/CD, automation, and keeping things shippable.
ML data pipelines: Interest in or exposure to ML data pipelines and robot learning.