2027 Internship Safety Engineer, Agentic Safety Case Assessment

Bedrock Robotics Inc • San Francisco, CA

45d

About The Position

Putting an autonomous 40-ton excavator on a live jobsite means being able to say, precisely, what we know about its safety and what that knowledge rests on. Today that argument lives in hazard analyses, requirements, tests and logs, linked by hand.
As a Safety Engineering intern, you’ll explore how formal methods and LLM-assisted proving can strengthen that argument. Working at the intersection of formal methods, statistics, and field robotics, you’ll formalize a narrowly scoped safety claim, verify a checker in Lean, and evaluate it on real fleet data, with explicit assumptions and auditable results.

Requirements

• Hands-on experience with Lean and mathlib experience, particularly real analysis or probability.
• Working knowledge of probability and statistics (distributions, confidence intervals, tail bounds).
• Familiarity with knowledge graphs or structured document analysis
• Ability to translate informal claims into precise specifications and identify their assumptions
• Experience evaluating LLM-generated code or proofs.
• Background in autonomous vehicles, heavy equipment, or other safety-critical robotics.

Nice To Haves

• Exposure to safety or assurance-case practice (GSN, UL 4600, ISO 13849, IEC 61508) or runtime verification.
• Experience with Rust verification tools (Aeneas, Verus, Kani) or differential/property-based testing.
• Familiarity with knowledge graphs or structured document analysis
• Background in autonomous vehicles, heavy equipment, or other safety-critical robotics

Responsibilities

• Become fluent in Bedrock's safety case and trace safety claims through the hazards, requirements, mitigations, and evidence that support them
• Build an end-to-end demonstration for one claim: define the property and assumptions, implement a checker in Lean, prove its correctness against the specification, and run it on fleet data
• Develop checks for assumptions that can be monitored in fleet data, and document those that require other evidence
• Build audit tooling for LLM-assisted proving: record proof dependencies, document toolchain trust assumptions, and reject proof placeholders or unapproved axioms in automated builds
• Work with Safety, Systems, and Autonomy teams to review formal specifications and turn findings into safety case improvements
• Document the results, limitations, and recommendations for adoption in our review process

What This Job Offers

• Career Level: Intern
• Education Level: No Education Listed