Senior AI/ML Engineer

General Motors • Sunnyvale, CA

18d • $170,600 - $261,300 • Hybrid

About The Position

The Senior AI/ML Engineer will work on GM’s Autonomous Vehicle organization’s Compression and Parity team, which develops model optimization and parity-analysis tools for models deployed in vehicles. The role owns model numerics: measuring, bounding, and reasoning about numerical behavior and turning analysis into deployment decisions. It connects differences caused by quantization, compilation, and precision reduction to downstream driving behavior, supported by quantitative evidence and investigation of the mechanisms behind observed differences.

The engineer will establish rigorous equivalence criteria for optimized implementations, develop sensitivity and robustness analysis tooling, and create diagnostics for training dynamics and numerical instability. These tools must be suitable for repeated use across checkpoints and production training and evaluation loops, with negligible throughput cost and no OOM risk. This position is categorized as hybrid, with the selected candidate expected to report to a specific location at least three times a week or at another frequency set by their manager. Travel is less than 25%; relocation benefits are available to qualifying candidates under company policy.

Requirements

• Working command of numerical analysis and matrix theory, including Jacobian and Hessian estimation, spectral properties, conditioning, and floating-point error analysis.
• Ability to reason about estimator variance and when it matters.
• Strong mental model of neural network training, including loss landscapes, gradient and error propagation, optimizer dynamics, and training failure mechanisms.
• Adversarial instinct for numerical edge cases, including denormals, extreme dynamic range, catastrophic cancellation, degenerate shapes, and accumulation-order effects.
• High proficiency in PyTorch and Python.
• Track record of building mathematically defensible analytical tools that are fast enough for production training and evaluation loops.
• Judgment to make analysis actionable, select quantities that answer the question being asked, and trace anomalous values to their source operations.
• Bachelor's, Master's, or PhD in Applied Mathematics, Control, Physics, Computer Science, Data Science, or a closely related quantitative field.

Nice To Haves

• Hands-on experience debugging large-scale training runs, diagnosing loss spikes, resolving numerical divergence, and investigating training dynamics.
• Experience with distributed training beyond DDP, such as FSDP, Megatron-LM, DeepSpeed, or 3D parallelism, particularly low-overhead observability over sharded parameter and gradient state.
• Experience in quantization, compiler toolchains, or inference-time numerical parity.
• Experience building or scaling evaluation pipelines and formulating metrics for AV or ADAS systems.
• Published or applied work in model robustness, OOD generalization, or adversarial/perturbation analysis.

Responsibilities

• Define what equivalence means for optimized implementations and design inputs that can expose violations.
• Connect numerical differences from quantization, compilation, and precision reduction to downstream driving behavior using open-loop metrics and closed-loop outcomes, and identify the mechanisms behind the relationship.
• Build Jacobian- and Hessian-based sensitivity and robustness analysis tooling for weight and input perturbations, including out-of-distribution inputs, and make it usable repeatedly across checkpoints by other engineers.
• Design diagnostics to detect and root-cause training instabilities, including vanishing or exploding gradients, loss spikes, and silent divergence.
• Develop decompositions of gradient and update trajectories into loss-descent and oscillatory components under modern training schedules such as WSD.
• Ensure metric computation, gradient decomposition, and diagnostic logging can run in distributed training jobs with negligible throughput cost and no OOM risk.
• Trace parity failures to the operations that caused divergence and explain why they occurred.
• Turn numerical analysis into actionable information for engineers making deployment decisions.

Benefits

• Medical, dental, and vision benefits
• Health Savings Account
• Flexible Spending Accounts
• Retirement savings plan
• Sickness and accident benefits
• Life insurance
• Paid vacation and holidays
• Incentive pay program with payouts based on company performance, job level, and individual performance
• Relocation benefits for qualifying candidates under company policy

What This Job Offers

• Job Type: Full-time
• Career Level: Senior