Senior Software Engineer - Machine Learning

📋 Description

• • Own the end-to-end design and development of machine-learning and reinforcement-learning models that simulate human-like road users—cars, cyclists, pedestrians—inside GM’s ultra-realistic AV simulation environment. Your models will become the “AI actors” that teach our autonomous vehicles how to handle the unpredictable dynamics of real-world traffic.
• • Architect and optimize training loops that can spin up thousands of concurrent RL experiments on GPU clusters, cutting iteration time from days to hours while preserving statistical fidelity. You will profile CPU/GPU utilization, schedule jobs, and tune hyper-parameters so that every cycle delivers maximum learning value.
• • Build the data pipelines and MLOps infrastructure that move petabytes of sensor logs, scenario labels, and model checkpoints from on-road fleet to cloud storage to training farm and back to simulation. Expect to work with TensorFlow, PyTorch, Ray, Kubernetes, and custom C++ services to guarantee fault-tolerant, low-latency delivery.
• • Define metrics that quantify realism, coverage, and safety impact of generated scenarios. You will partner with Safety & Validation teams to translate regulatory requirements into measurable objectives, then surface dashboards that allow leadership to see at a glance how each new model improves AV disengagement rates.
• • Integrate your RL policies with the broader Simulation stack, collaborating daily with Behavior, Perception, and Maps teams. You will write clean APIs, version contracts, and regression tests so that new perception models or map updates never break downstream agents.
• • Champion software-engineering best practices across the ML org: code reviews, unit/integration tests, CI/CD, reproducible notebooks, and experiment tracking. Your mentorship will level-up junior engineers and create a culture where research ideas reach production without accruing technical debt.
• • Rapidly prototype Generative-AI approaches—think GPT-style sequence models—to create novel edge-case traffic scenarios that human designers would never imagine. You will run A/B experiments that prove these synthetic scenes accelerate policy convergence and reduce real-world disengagements.
• • Contribute to GM’s open-source simulation projects and publish at top-tier ML/Robotics conferences. Your thought leadership will attract world-class talent and reinforce GM’s reputation as the place where AV research meets industrial-scale impact.
• • Balance deep technical dives with broad system thinking: one day you may be optimizing CUDA kernels for a new attention mechanism, the next you may be presenting a roadmap that spans data ingestion, model training, and on-car deployment.
• • Work fully remote from anywhere in the United States, while staying tightly connected via Slack, Miro, and immersive VR stand-ups. If you live within 50 miles of Atlanta, Austin, Detroit, Warren, Milford, or Mountain View, plan to visit your home site three times a week for white-boarding and hardware-in-the-loop sessions.