Autonomy Engineer - Deep Learning Model Acceleration

Skydio · Defense · San Mateo, CA +1 · R&D

The role focuses on building and scaling infrastructure for deep learning and AI efforts, specifically for high-performance inference of computer vision workloads on various hardware platforms. It involves profiling, optimizing, and deploying models, as well as creating MLOps workflows and potentially implementing GPU kernels. The goal is to accelerate progress in intelligent mobile robots by leveraging visual data for semantic and geometric understanding.

What you'd actually do

  1. Develop solutions for high-performance deep learning inference for CV workloads that can deliver high throughput and low latency on different hardware platforms
  2. Profile CV and Vision Language Models (VLMs) to analyze performance, identify bottlenecks and acceleration/optimization opportunities and improve power efficiency of deep learning inference workloads
  3. Design and implement end to end MLOps workflows for model deployment, monitoring, and re-training
  4. Utilize advanced Machine Learning knowledge to leverage training or runtime frameworks or model efficiency tools to improve system performance
  5. Create new methods for improving training efficiency

Skills

Required

  • MLOps
  • ML inference acceleration/optimization
  • edge deployment
  • DL fundamentals
  • CV fundamentals
  • image processing
  • video processing
  • ML pipelines
  • data preparation
  • model training
  • model deployment
  • monitoring
  • ML frameworks and libraries
  • software lifecycle (architecture, development, testing, deployment, monitoring)

Nice to have

  • Vision Language Models (VLMs)
  • GPU kernels
  • SDK development
  • security and compliance requirements in ML infrastructure

What the JD emphasized

  • high-performance deep learning inference
  • MLOps
  • ML inference acceleration/optimization
  • edge deployment
  • CV workloads
  • model deployment
  • model acceleration
  • optimization

Other signals

  • deep learning inference
  • MLOps
  • model acceleration
  • optimization
  • edge deployment