CV/ML Platform Engineer

What You'll Do

• Deploy and operate Kubernetes clusters on bare-metal infrastructure hosting 130+ NVIDIA GPUs, with hybrid burst capability to AWS for scalable compute and storage workloads.
• Manage NVIDIA GPU clusters for ML training.
• Own the ACS CV/ML CI/CD pipeline.
• Improve and maintain core ML infrastructure, such as model registration and versioning, experiment tracking, and model and data provenance tracking.
• Improve and maintain ML model testing, performance analysis, and reporting tools.
• Automate repetitive model training and testing tasks to increase developer velocity.
• Work with Software Team Platform Engineers to ensure efficient coordination and minimal duplication between CV/ML infrastructure and wider Software infrastructure.
• Collaborate with the Software Team to automate the optimization of models (TensorRT/quantization) for deployment on NVIDIA Jetson and other edge hardware.

Required Technical Skills

• 2+ years of experience in Platform Engineering or DevOps/MLOps.
• Strong programming skills are required for automating ML lifecycles and building custom CLI tools for CV engineers.
• Hands-on experience with NVIDIA GPU infrastructure, including managing CUDA libraries and development environments, GPU Operator, device plugins, and scheduling (MIG, Volcano, or fractional GPU sharing).
• Experience implementing and maintaining MLOps platforms such as Kubeflow, MLflow, Weights & Biases (W&B), or DVC for experiment tracking and model versioning.
• Familiarity with high-performance storage solutions (e.g., MinIO, WEKA, or Ceph) and data orchestration tools capable of handling terabytes of video/image data.
• Proven track record building CI/CD pipelines that include automated model validation, performance benchmarking, and artifact management for both cloud and edge targets.
• Experience with model optimization toolchains, including TensorRT, ONNX, and quantization techniques, specifically for cross-compilation to ARM targets like NVIDIA Jetson.
• Proficiency with observability stacks (ELK, Prometheus/Grafana) adapted for ML, including monitoring GPU health, training throughput, and model inference metrics.
• Strong Linux systems knowledge (Debian/Ubuntu), including networking for high-throughput data, storage, and security hardening for defense-grade production environments.