Neuro-symbolic AI Engineer

Common Sense Robotics builds intelligent robot systems for manufacturing environments where the cost of failure is high: aerospace, automotive, and other industries where every robot decision must be traceable, auditable, and explainable. Our clients include Safran and Audi. To deliver on this we are building the "Task Execution System" (TES): a white-box, ontology-driven robot operating system designed from the ground up as an explainable complement to ROS and end-to-end foundation model approaches. Where most of the field is working on opaque learned policies, we are making the orchestration system that can integrate deep and reinforcement learning policies with symbolic, inspectable reasoning.

The Role:

Robots in regulated environments only get to act on what they can prove they understand. Process specifications, work instructions, safety constraints, and quality requirements all live in dense technical documents. Before any of this can drive a robot, it has to be lifted out of unstructured text and into a queryable, symbolic representation that downstream reasoning layers can rely on. You will build the pipeline that makes this possible. Concretely, you will design and ship the agentic neuro symbolic system that ingests technical documents and produces a structured knowledge graph aligned with our ontology. The agents you build will not just extract text: they will call tools that validate, cross reference, and ground their outputs against the symbolic model, so that what enters the knowledge graph carries the guarantees that regulated industries require.

Concretely, you will:

• Design and build the document ingestion pipeline end to end: from raw technical documents to a validated, queryable knowledge graph.
• Architect the agentic layer: which agents exist, what tools they have access to, how tool calls are designed, and how outputs are verified before they are committed.
• Work at the seam between LLMs and symbolic AI: using language models for what they are good at (reading, extracting, normalizing) while leaning on the symbolic layer for what LLMs are bad at (consistency, auditability, guarantees).
• Collaborate closely with the team building the robot reasoning layer that consumes your output, and with the engineers shaping the underlying ontology.
• Engage with domain experts to understand how technical documents are actually structured in our customers' industries, and translate that into ingestion strategies that hold up in production.

You likely have:

• Two to three years of relevant professional experience in AI or automation solution deployment.
• Hands on experience with LLM based agentic systems: not just prompting, but designing agents, tools, tool calls, and the control flow around them.
• Solid prompt engineering instincts and an understanding of where LLMs are reliable and where they are not.
• Experience with knowledge representation and graph structured data: ontologies, SPARQL or GQL/Cypher, GraphRAG or comparable retrieval patterns.
• The ability to design clean, maintainable pipelines with an eye for long term architecture rather than one off scripts.
• Comfort collaborating with infrastructure, cloud, and data teams to deliver integrated outcomes.
• Strong programming skills in Python, with working knowledge of C++ and graph query languages.

If this matches your expertise and ambitions, we look forward to your application.

How to apply: We don't review generic CVs. Instead, send us a short application document, written specifically for this role, explaining why you're a good fit and proposing concretely how you would approach one or two of the challenges described above. email: [email protected]