We’re building an AI system that automatically recognizes and maps CAN signals in cars to speed up porting vehicles to openpilot-like autonomy stacks.
Instead of manually reverse-engineering every car, we collect CAN logs, train models, and let AI suggest signals (speed, steering, brake, gas, etc.) that engineers only confirm.
This is not a CRUD web app.
This is a real data/ML system handling large logs, background jobs, inference, and user workflows.
What you’ll build
You will design and implement the backend for:
- uploading and storing large CAN logs
- parsing and preprocessing data
- background processing pipelines
- model inference APIs (serving predictions)
- labeling workflows (human-in-the-loop training)
- dataset/version management
- integration with our online Cabana-like web tool
- exporting signal definitions (DBC/openDBC style)
In short:
You build the data + API + ML infrastructure layer.
Tech stack (flexible but preferred)
- Python (FastAPI) or Node.js (NestJS)
- PostgreSQL
- S3/MinIO object storage
- background jobs (Celery / Redis / queues)
- Docker
- Linux
Nice to have:
- WebSockets
- ClickHouse / time-series DB
- basic MLOps knowledge
- experience with large file processing
- experience with data pipelines
Requirements
Must have:
- solid backend fundamentals
- API design skills
- database knowledge (SQL, schema design)
- async/background processing
- Docker
- ability to work independently
Good fit if you:
- like systems engineering more than frontend/UI
- enjoy working with real hardware/data
- want to build something used in real vehicles
- prefer practical engineering over “AI hype”
Not required:
- automotive knowledge
- CAN knowledge
- deep ML knowledge
You will learn those on the job.
What makes this interesting
This is NOT:
- a typical web shop or dashboard
This IS:
- automotive reverse engineering
- real CAN buses and hardware
- ML + data engineering
- building tools used to port real cars
- solving non-trivial technical problems
You’ll work closely with:
- ML engineer
- CAN expert
- embedded/hardware team
Very hands-on, very practical.
We offer
- part-time or full-time
- student-friendly
- flexible hours
- remote or hybrid
- real ownership (not tickets from Jira hell)
- direct impact on product
- modern stack, no legacy corporate nonsense
Nice bonus if you have
- built any backend projects yourself
- worked with data pipelines
- played with Raspberry Pi / hardware / robotics
- contributed to open source
- built tools, not just websites