Research Engineer - Ai/rl Infrastructure

Applied Intuition Applied Intuition · Robotics · Sunnyvale, CA · AI Research

Research Engineer focused on building and operating large-scale ML training and evaluation infrastructure for physical AI systems, including autonomous driving and robotics. The role involves orchestrating GPU clusters, developing benchmarking and data pipelines, and enabling distributed training across cloud environments.

What you'd actually do

  1. Design and build training and evaluation infrastructure to support our current AI research directions, orchestrating massive GPU clusters to process PBs of multimodal sensor data
  2. Build robust benchmarking, continuous evaluation, and regression tracking systems to measure model performance across diverse, long-tail real-world driving distributions
  3. Develop large-scale data sampling, dataset generation, and advanced data curation pipelines, leveraging state-of-the-art AI models to power a closed-loop data flywheel
  4. Enable high-throughput distributed training across heterogeneous cloud environments, focusing on reliability, efficiency, and cost-aware scaling
  5. Collaborate closely with AI research, autonomy, and platform teams to translate cutting-edge research into production-ready systems

Skills

Required

  • Experience building and operating production-grade software systems across the full machine learning lifecycle, including training, evaluation, data, and deployment
  • Opinions about building a company-wide platform for ML training, evaluation, and deployment
  • Experience with performance engineering and compute acceleration for large-scale ML training, including profiling, bottleneck analysis, and optimization
  • Strong systems-level debugging skills to diagnose and resolve issues in large-scale distributed training, spanning model code, data pipelines, runtimes, and cluster infrastructure
  • Deep familiarity with the open-source ML and systems ecosystem, with judgment on when to adopt open source versus build in-house
  • Pytorch
  • CUDA
  • Ray
  • Flyte
  • K8s

Nice to have

  • Industry experience on relevant topics (self-driving application preferred)

What the JD emphasized

  • training and evaluation infrastructure
  • benchmarking, continuous evaluation
  • large-scale data sampling, dataset generation, and advanced data curation pipelines
  • high-throughput distributed training
  • production-grade software systems across the full machine learning lifecycle
  • company-wide platform for ML training, evaluation, and deployment
  • performance engineering and compute acceleration for large-scale ML training
  • systems-level debugging skills to diagnose and resolve issues in large-scale distributed training

Other signals

  • large-scale ML systems
  • physical AI
  • autonomous driving
  • robotic generalist
  • training and evaluation infrastructure
  • multimodal sensor data
  • distributed training
  • heterogeneous cloud environments