Inactive
O

OpenAI

Builds AI models and the ChatGPT assistant

ML Research Engineer - Hardware Codesign

AI ResearchPosted on Jan 13
Full-timeSan Francisco, CA, USA$185k - $455k /yr

About the Job

Requirements

  • Strong Python and C++ or Rust programming skills.
  • Working knowledge of PyTorch or JAX.
  • Practical understanding of floating-point numerics, reduced-precision machine-learning tradeoffs, and model quantization.
  • Deep understanding of transformer models, transformer rooflines, and the tradeoffs of sharded training and inference in large-scale machine-learning systems.
  • Strong cross-functional communication across machine-learning researchers and hardware engineers.
  • Ability to turn ambiguous early-incubation ideas into concrete areas for progress.
  • A track record of producing high-quality technical output and iterating prototypes when requirements are unclear.

Responsibilities

  • Build on the roofline simulator to track evolving workloads and deliver analyses that quantify the impact of system architecture decisions and support technology pathfinding.
  • Debug gaps between performance simulation and real measurements, and communicate root causes, bottlenecks, and invalid assumptions.
  • Write emulation kernels for low-precision numerics and lossy compression schemes, and provide Research with information to trade efficiency against model quality.
  • Prototype numerics modules by pushing register-transfer level designs through synthesis, hand off novel numerics, and occasionally own a register-transfer level module end-to-end.
  • Pull in new machine-learning workloads, prototype them with rooflines or functional simulation, and drive initial evaluation of new opportunities or risks.
  • Understand the end-to-end path from machine-learning science to hardware optimization and divide the objective into near-term deliverables.
  • Build collaborations across teams with different goals and areas of expertise to keep progress unblocked.
  • Communicate design tradeoffs with explicit assumptions and confidence levels, and produce evidence that enables execution.

Desired Qualifications

  • Experience in large machine learning codebases.
  • Experience writing register-transfer level designs, especially for floating-point logic, and understanding of power, performance, and area tradeoffs.
  • Experience with Triton, CUDA, or similar kernel programming.
  • Experience in a related technical area is a plus.

About the Company

OpenAI

Builds AI models and the ChatGPT assistant

OpenAI researches and builds artificial intelligence models and turns them into products for people, developers, and businesses, with a stated mission of making sure artificial general intelligence benefits all of humanity. Its launch of ChatGPT in 2022 brought generative AI into everyday use. People use ChatGPT to ask questions, write, code, and analyze files, developers build OpenAI's models into their own apps through its API, and its Codex agent works through coding tasks on its own. OpenAI earns money from ChatGPT subscriptions, business plans, and API usage. Sam Altman, Elon Musk, and others co-founded it in 2015; it is based in San Francisco and overseen by a nonprofit, the OpenAI Foundation.

Company size
1,001-5,000
Company stage
Late Stage VC
Total funding
$207.5B
Headquarters
San Francisco, CA
Founded
2015

Recently Posted Jobs

Explore more opportunities at OpenAI

See more jobs