Software Developer in Berlin

AURYAL is building the supervision layer for physical AI. We turn human cognition into dense training signal for robots, autonomous systems and other embodied agents. Our platform turns multimodal data from simulators, sensors and cohorts into training and evaluation workloads for reinforcement learning and vision-language-action (VLA) models.

We are looking for a Software Developer who wants to build the core systems that make this possible.

Your role

You will work closely with our architects, ML engineers and neurotech team to implement the services and pipelines that power our product. This is an execution-focused role: design clean solutions, write high-quality code, and help us get features over the line.

You will:

• Implement backend services and APIs that ingest and synchronise multimodal data (video, telemetry, cognition signals, annotations).
• Build and maintain data processing pipelines that transform raw partner and cohort data into ML-ready datasets for RL and VLA training.
• Integrate our platform with external simulators, robotics environments and log-replay systems.
• Implement experiment and evaluation tooling (e.g. running batch evaluations, exporting metrics and reports for partners).
• Contribute to internal tools that help us run and monitor training jobs on GPU infrastructure.
• Write tests, improve observability and keep our systems reliable as we move from PoCs to production deployments.
• Participate in design reviews, code reviews and joint debugging sessions; propose pragmatic improvements rather than abstract perfection.

What we are looking for

• 3+ years of professional experience as a software engineer.
• Strong programming skills in Python or another modern language (e.g. Go, TypeScript/Node) and willingness to work heavily in Python where it touches ML.
• Solid experience with backend development: REST/gRPC APIs, authentication, configuration, logging, testing.
• Experience with databases and data modelling (SQL and/or NoSQL) and comfort working with large datasets.
• Familiarity with containerisation (Docker) and at least one cloud platform (AWS, GCP or Azure).
• Ability to take a loosely specified requirement, clarify it, and ship a maintainable implementation.
• Comfortable working in an environment where research code and production code coexist, and where requirements evolve as we learn from partners.
• Good written and spoken English.

Nice to have

• Experience building data or ML pipelines (e.g. Airflow, Prefect, Luigi, custom solutions).
• Exposure to reinforcement learning or imitation learning setups (RLlib, CleanRL, custom PyTorch/JAX training loops).
• Experience integrating with robotics or autonomous driving stacks: ROS2, Isaac Sim, Mujoco, Gazebo, CARLA or similar.
• Experience with multimodal or VLA models (vision-language / perception-action systems) and their data requirements.
• Knowledge of monitoring/observability stacks (Prometheus, Grafana, OpenTelemetry) and basic SRE practices.

What we offer

• The chance to work on real problems at the intersection of cognition, robotics and reinforcement learning.
• Close collaboration with an experienced founder team and experts in AI and neurotechnology.
• Ownership over meaningful parts of the codebase, not just tickets on a board.
• Competitive compensation (salary + equity) and flexible work arrangements centred around Berlin.

#J-18808-Ljbffr