AI Platform Evaluation Software Engineer - Autonomous Vehicles - General Motors - Milford, MI (+1 other)

Role Overview

As a member of the core AV software reliability team , you will be responsible for ensuring safe, stable, and scalable Autonomous Vehicle (AV) software releases by turning failures into actionable engineering insights at scale.

This is a software-first, platform-focused role. You will work primarily across the AV platform software stack (frameworks, runtime, services, orchestration, data pipelines) and its interaction with vehicle hardware and compute —not cloud infrastructure or hardware design.

The mission of this role is to:
• Improve learning velocity from failures
• Reduce reliability escapes
• Increase confidence in production readiness

You will do this through intelligent triage, deep software debugging, and AI-assisted failure analysis across simulation, CI, HIL, SIL, and on-road environments , ensuring that failures are:
• Correctly detected and interpreted
• Consistently categorized and de-duplicated
• Rapidly mapped to the right owners and solution space

You will collaborate closely with AV software engineers, ML engineers, systems engineers, test platform owners, and release/safety stakeholders to ensure reliability signals directly influence engineering priorities and release decisions .

If you are passionate about software reliability, failure analysis, and building AI-driven systems that help organizations learn faster from complex ML-based AV software, this role is for you.

Key Responsibilities
• Own the AV software reliability triage framework for the on-vehicle / AV platform stack, defining how failures from simulation, CI, HIL/SIL, and on-road validation are detected, grouped, and escalated into actionable tickets and insights.
• Perform deep debugging and root-cause analysis across:
• AV platform and framework code
• Perception / planning / control software integrations
• ML pipelines and model rollouts
• Vehicle compute and hardware interfaces (sensors, ECUs, networks)
Connecting failure symptoms (logs, time-series, traces)