Data Engineer, Robot Learning Data Platform | Munich | Early-stage Physical AI

Ai Futures; München; On-site; EUR 90,000 - 135,000; Full time

Job summary

Ai Futures is building the data foundation for next-generation industrial robots. We are seeking a data platform lead to own end-to-end data flow from edge logging to labeled, anonymized datasets.

You will design the datalake, ensure data quality, and enable ML teams to access training data with traceability. You will work on privacy-preserving ingestion, schema migrations, and scalable pipelines, collaborating closely with ML engineers and robotics researchers in Munich, Germany.

Qualifications

• Strong Python programming.
• Production experience with relational and non-relational databases.
• Experience building data lakes, ingestion, storage and labeling tooling.
• Familiarity with data serialization formats (protobuf, Avro).
• Experience with schema design and migrations (e.g. Alembic).

Responsibilities

• Own logging correctness across video, audio, events and telemetry, including timestamp discipline.
• Build and run the datalake: ingest, storage and labeling tooling.
• Make data self-serve so ML engineers can assemble datasets independently.
• Verify anonymization at ingest.
• Build the export and encoding pipeline for robot-learning research.

Skills

Python

Databases (SQL & NoSQL)

Data pipelines

Data serialization (protobuf, Avro)

Schema design & migrations

Alembic

Tools

Label Studio

FiftyOne

Job description

We're partnering with an early-stage Physical AI company that's building the data foundation for the next generation of industrial robots.

Their product is an AI copilot for smart glasses that guides factory workers through assembly and inspection in real time. It's already live with leading automotive manufacturers. Every shift also produces something rare: real multimodal recordings from the factory floor, which become training data for robot learning.

They're looking for someone to own that data flywheel end-to-end, from the first byte logged on an edge device to a labeled, queryable, anonymized dataset that an ML engineer can use without asking anyone for help.

What you'll do

• Own logging correctness across video, audio, events and telemetry, including timestamp discipline
• Build and run the datalake: ingest, storage and labeling tooling
• Make data self-serve, so ML engineers assemble datasets on their own and every deployed model traces back to its exact training data
• Verify anonymization at ingest, because privacy is verified, never assumed
• Build the export and encoding pipeline that feeds the company's robot-learning research

What you bring

• Strong Python
• Production experience with relational and non-relational databases
• Object storage and data pipelines at real scale
• Data serialization formats and their evolution (protobuf, Avro)
• Schema design and migrations (e.g. Alembic)

Nice to have

• C++ or Rust alongside Python
• ML labeling and curation tooling (Label Studio, FiftyOne)

Why this role

• Sole ownership of the foundation that both client delivery and robot-learning research stand on
• Schema and lineage decisions that compound for years, made by you, early
• Short feedback loops: you build for the ML team and see the impact on their work immediately
• Data most teams never see: real egocentric recordings from factory floors
• Join early and build the platform alongside the team, rather than inherit a finished one