Distributed LLM Inference Engineer

Anyscale Anyscale · Data AI · San Francisco, CA · Engineering

Anyscale is seeking a Distributed LLM Inference Engineer to build and optimize systems for large-scale LLM inference, integrating with Ray and open-source projects like vLLM. The role focuses on achieving high throughput and low latency for batch and online inference, contributing to Anyscale's market-leading AI infrastructure.

What you'd actually do

  1. Iterate very quickly with product teams to ship the end to end solutions for Batch and Online inference at high scale which will be used by open-source Ray users and customers of Anyscale
  2. Work across the stack integrating Ray Data and LLM engine providing optimizations achieving low cost solutions for large scale ML inference
  3. Integrate with Open source software like vLLM, work closely with the community to adopt these techniques in Anyscale solutions, and also contribute improvements to open source
  4. Follow the latest state-of-the-art in the open source and the research community, implementing and extending best practices

Skills

Required

  • Familiarity with running ML inference at large scale with high throughput and low latency
  • Familiarity with deep learning and deep learning frameworks (e.g. PyTorch)
  • Solid understanding of distributed systems, ML inference challenges

Nice to have

  • ML Systems knowledge
  • Experience using Ray
  • Work closely with community on LLM engines like vLLM, TensorRT-LLM
  • Contributions to deep learning frameworks (PyTorch, TensorFlow)
  • Contributions to deep learning compilers (Triton, TVM, MLIR)
  • Prior experience working on GPUs / CUDA

What the JD emphasized

  • high scale
  • low latency
  • high throughput

Other signals

  • LLM inference
  • distributed systems
  • high scale
  • low latency
  • Ray ecosystem