Dyna Robotics

ML Infrastructure Engineer, Training

3mo ago
220000 –320000 USD / yearUSASenior
Dyna Robotics

ML Infrastructure Engineer, Training

3mo ago
220000 –320000 USD / yearUSASeniorgpukubernetesslurmdistributed trainingmemory optimizationshardingactivation checkpointing

Architect and build training infrastructure to optimize multi-cloud GPU fleet usage for AI-driven robotics research.

Nice to have

  • Experience with Robotics Data Formats (MCAP, Protobuf) or multimodal models (VLAs).
  • Deep ML systems experience: custom kernels (Triton), compilers, or runtime optimization.
  • Experience as a founding or early-stage infrastructure hire.
  • At Dyna Robotics , we build technology for the real world, which requires a team as diverse as the environments our robots inhabit. We are an equal opportunity employer committed to technical rigor and mutual respect.
  • Don’t let a checklist stop you. Data shows that underrepresented groups often only apply if they meet 100% of the criteria. We value problem-solving and grit over keyword matching. If you’re passionate about the intersection of geometry and robotics, we want to hear from you—even if you don't check every box.

Other

  • Dyna Robotics makes general-purpose robots powered by a proprietary embodied AI foundation model that generalizes and self-improves across varied environments with commercial-grade performance. Dyna's robots have been deployed at customers across multiple industries. Its frontier model has the top generalization and performance in the industry.
  • Dyna Robotics was founded by repeat founders Lindon Gao and York Yang, who sold Caper AI for $350 million, and former DeepMind research scientist Jason Ma. The company has raised over $140M, backed by top investors, including CRV and First Round. We're positioned to redefine the landscape of robotic automation.
  • As a ML Training Infrastructure Engineer, you will architect and build the systems that turn our multi-cloud GPU fleet into a training engine our researchers love. Your charter is singular and broad: own training infrastructure end-to-end so that every GPU is busy, every run is reproducible, and every researcher's next experiment is one command away.
  • Scale Distributed Training: Architect and own the infrastructure for large-scale GPU clusters. You’ll implement sharding, activation checkpointing, and memory optimization (ZeRO, FSDP) to enable the training of massive multimodal models.
  • Optimize Researcher Ergonomics: Build a research codebase and job scheduling system (Kubernetes/SLURM) that prioritizes fast iteration, automated retries, and seamless failure recovery.
  • High-Performance Data Handling: Design high-throughput pipelines to ingest and transform terabytes of multimodal robot data (video, proprioception, 3D signals), ensuring dataloaders never starve the GPUs.
  • Production Inference: Build low-latency inference pipelines for real-time robot control. You’ll apply quantization, distillation, and model compilation (TensorRT, Triton) to move models from the lab to the physical world.
  • Deep Systems Profiling: Dive into the weeds of GPU utilization, I/O bottlenecks, and memory fragmentation to squeeze every bit of performance out of our expanding compute fleet.
  • 7+ Years of Engineering: With a track record of leading technical projects in high-performance computing (HPC) or ML infrastructure.
  • ML Systems Mastery: Deep experience with PyTorch and distributed training frameworks (DeepSpeed, Accelerate). You understand the nuances of mixed precision and gradient accumulation.
  • Infrastructure Expertise: Hands-on experience managing cloud GPU environments (GCP/AWS) and container orchestration (Kubernetes).
  • Low-Level Intuition: A fundamental understanding of distributed systems, including race conditions, memory management, and NCCL/inter-node communication.
  • Ownership Mindset: You don't just "deploy" code; you design, build, and operate systems end-to-end to unblock fast-moving research.