Software Engineer, Data Infrastructure

The role

You will build and operate pipelines that turn raw sensor and demonstration data into training-ready datasets. You will work on ingestion, processing, quality filtering, and dataloaders for large-scale multimodal datasets.

What you will do

• Build and scale ingestion pipelines for high-volume, high-dimensional sensor data.
• Build and maintain data quality and curation systems.
• Contribute to storage and retrieval design for large-scale multimodal datasets.
• Build and operate distributed processing infrastructure.
• Debug production issues in live data pipelines.
• Work with modeling/research partners to translate needs into working pipelines.
• Participate in design and code review across the data infrastructure stack.

Required qualifications

• 2+ years of software engineering experience, with some exposure to data pipelines, data engineering, or backend systems.
• Strong programming fundamentals in Python, with the ability to write performant, production-grade data processing code.
• Experience with at least one distributed data processing framework (e.g., Spark, Ray, Dask, or Flink), or strong fundamentals and willingness to ramp up quickly.
• Familiarity with data storage concepts — object storage, data lake table formats, and warehouse vs. lake tradeoffs.
• Bias for ownership: you've taken features or systems from prototype to production.
• Clear communicator who collaborates well with research/modeling partners and more senior teammates.

Nice to have

• Experience with workflow orchestration tools (e.g., Airflow, Dagster, Prefect)
• Experience with streaming ingestion for high-volume, near-real-time data
• Experience building automated data quality/curation systems (statistical filtering, anomaly detection, deduplication, or using ML models in an annotation/filtering pipeline)
• Experience with robotics- or embodied-AI-specific data (multi-embodiment datasets, teleoperation/demonstration data, egocentric video)

About the company

Mind Robotics is building Physical AI for real-world industrial deployment, starting with the factory floor. They believe the hardest problems in AI are solved when researchers and engineers are hands-on with the physical world every day.