Lead Test Engineer AV Safety Assurance

Avride builds autonomous driving technology for vehicles and delivery robots, and tests it on public roads in Austin. Before anything reaches those roads it goes through our QA team: checks on every pull request, a daily validation of main, and release-candidate qualification over a large set of recorded scenes. Every one of those gates produces a verdict, and the safety part of that verdict is the one people act on.

You will own how safety is tested, judged and reported across our QA team. Our Head of Safety owns the company's safety framework and the safety case built on top of it. You own the layer beneath: the gates, the criteria and the reports that turn day-to-day testing into evidence that argument can rest on.

The hard part of this is not the collisions. Those are obvious, and nobody argues about them. It is everything short of a collision  the near miss, the late brake, the hesitation at an intersection that looked wrong to one engineer and fine to another where the judgement has to be made and then made the same way again next week, by someone else, on a different run. Turning that into criteria a whole team applies consistently is the craft of this role.

You will lead through process and through people who do not report to you.

• Own the safety gates. What counts as a safety-relevant finding, how it is classified and escalated, the criteria that hold a merge or a release, and how findings are tracked to closure  with the recurring ones feeding back into the criteria.
• Own the acceptance criteria. Decide what we measure at each gate from collisions and near misses to right-of-way and hard braking where the thresholds sit, and move them as the technology changes.
• Own the safety report. Define the format the whole QA team writes to: comparable between runs, explicit about what was not covered, and short enough to be read before a decision rather than after it.
• Coach the QA team. Run calibration sessions, review reports, and teach the reasoning behind the criteria, so that safety judgement is something the team applies on its own rather than something you add at review.
• Be the safety voice in the release decision. Releases are decided by a review that brings together engineering, safety and operations. Your part is the one nobody else can supply: what the testing actually showed and what is still unknown  put plainly enough that the decision is made with its eyes open.
• Connect hazards to evidence. Take hazard analysis output  STPA, FMEA, HAZOP  through to what our scenario set has to contain, through to results, through to something our safety case can stand on.
• Work with the safety organization. Translate the safety framework into testable requirements with our Head of Safety, and work with development and analytics on the tooling that makes those requirements measurable.
• Keep up with the rules that apply to us. Track the standards and regulation governing driverless operation in Texas and beyond, and bring what matters into how we test.

• 7+ years in testing, verification or validation, including ownership of something that outlived your involvement — a gate, a review, a reporting standard.
• Experience where being wrong had physical consequences. Autonomous vehicles, aviation, medical devices, rail, industrial automation, defence — the field matters far less than the habits it builds: claims backed by evidence, limits stated rather than glossed over, and a clear head about what is still unknown.
• Hazard-driven thinking. You start from what can go wrong and work forward to what would have to be tested to know — and you can work with the output of a hazard analysis rather than needing it translated for you.
• Comfort with residual risk. You can say what we do not know and how much of it we are accepting, without overstating it and without smoothing it away.
• Influence without authority. Much of this job is persuading engineers and QA colleagues to change how they work because your argument is good, not because you can require it.
• A habit of checking the number yourself. When a safety metric looks wrong, you go to the data behind it rather than accepting the dashboard.
• Fluent use of LLMs as a working tool — reading through standards, logs and results quickly, and drafting the documents this role produces.
• Excellent written English. Other people's decisions rest on your documents.

• Working knowledge of the standards that govern this work — ISO 21448 (SOTIF), ISO/TS 5083, UL 4600, ISO 26262, ISO 34501–34505 — or their equivalents from your own field, such as DO-178C, IEC 62304 and ISO 14971, EN 50716, or IEC 61508. Whoever takes this role will become fluent in the AV-specific ones; we do not expect you to arrive that way.
• Autonomous vehicles, ADAS, or another domain that operates without a human fallback.
• Direct contribution to a safety case, a Voluntary Safety Self-Assessment, or a regulatory submission.
• Experience defining on-road testing safety protocols and safety-driver procedures.
• Experience hiring engineers.
• Hands-on SQL and Python for digging into results yourself.