SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Atoms is building Physical AI—real-world robots for industries including food, mining, and transport. This role owns the safety-relevant scenarios used to validate autonomous systems before deployment.
You will define what makes a scenario safety-relevant, build and maintain the scenario catalog, and own the technical argument that coverage is sufficient. The work spans two distinct sources: log-based scenarios extracted from real fleet data (disengagements, interventions, near-misses) and NCAP-based scenarios from published consumer test protocols adapted for autonomous platforms.
Key responsibilities include:
• Scenario definition and taxonomy: establish criteria for safety relevance, criticality measures, and maintain coherent abstraction levels (functional, logical, concrete) as the catalog grows.
• Log-based scenario extraction: build pipelines to mine fleet data for safety-critical events, own criticality metrics (time-to-collision, post-encroachment time, margin to drivable envelope), and understand metric limitations.
• Crash-avoidance protocol suites: implement and maintain standardized test suites from NCAP and equivalent protocols (AEB car-to-car, AEB vulnerable road users, lane support, emergency steering), including test conditions and pass/fail criteria.
• Execution and evidence: define how scenarios run across resimulation, SIL/HIL, vehicle-in-the-loop, closed course, and on-road testing; own regression suites to prevent known failures from returning.
• Interfaces: translate triggering conditions and hazards from functional safety and field operations into scenario-based validation evidence that feeds the safety case.
This is a technical individual contributor role. You will build pipelines, write analysis, and your coverage position becomes part of the formal safety argument. The role is expected to leverage AI tooling—LLM-based systems and purpose-built tools for scenario mining, categorization, criticality assessment, and documentation—while maintaining rigorous human judgment about what constitutes valid coverage. You should be comfortable writing extraction pipelines, coverage analysis, and reporting tools yourself, and understanding the limits of simulation fidelity and evidence.
