Normal Computing
Normal Computing
51 – 200 Angajați
Informațiile disponibile despre companie menționează că Normal Computing a fost fondată în 2022, dar nu descriu produsele, serviciile, domeniul de activitate sau contextul angajărilor sale. Sunt necesare mai multe detalii specifice companiei pentru a oferi o prezentare utilă celor care caută un loc de muncă.

Compiler Frontend Engineer, LLM Architectures

Extend compiler support for emerging LLM architectures, from attention variants and MoE routing to quantization and numeric formats. Validate model bring-up on simulators and FPGA prototypes for Normal Computing’s in-memory AI hardware.

Descriere

  • Extend frontend and mid-level compiler support for emerging LLM architectures.
  • Implement attention variants, MoE routing schemes, and other architectures in the compiler.
  • Expand the IR, operations, and lowering passes to represent new architectural primitives accurately.
  • Maintain test coverage as the compiler evolves.
  • Represent quantization schemes and numeric formats while preserving model accuracy.
  • Monitor developments in LLM architectures and recommend priorities for future compiler support.
  • Pass lowered programs to backend compiler engineers for hardware mapping.
  • Work with the LLM serving engineer to move models from compiler support into serving.
  • Validate compiler bring-up with simulators and FPGA prototypes before silicon is available.

Cerințe

  • Hands-on experience developing and shipping ML compiler passes.
  • Strong familiarity with PyTorch internals, ideally including TorchDynamo, FX, and torch.export.
  • Working knowledge of modern LLM architectures, including attention mechanisms, MoE routing, and quantization schemes.
  • Comfort designing compiler infrastructure in a new codebase and extending an established one.
  • Comfort building software for evolving hardware, from simulation through bring-up.
  • Fluency in Python.
  • Familiarity with C++ or Rust.
  • Experience with IREE, XLA, or TVM is a plus.
  • Familiarity with JAX or ONNX is a plus.
  • Experience partitioning models across devices with tensor, pipeline, or expert parallelism is a plus.
  • Experience with hardware-software co-design is a plus.
  • Experience with non-standard or in-memory hardware targets is a plus.
  • Contributions to open-source compiler or ML infrastructure projects are a plus.

Beneficii

  • Research-adjacent work with the opportunity to publish on behalf of Normal Computing.
  • Work on pre-silicon hardware using simulators and FPGA prototypes.
  • Inclusive workplace with an Equal Employment Opportunity commitment.
  • Reasonable accommodations are available for individuals with disabilities.

Locuri de muncă similare

Commure

Senior Applied AI Software Engineer

Commure

Build LLM-powered revenue cycle automation for Commure’s healthcare AI platform. Develop dependable systems for claims, remittances, denials, appeals, and document extraction.

Deschide
Commure

Senior Software Engineer, Claims Intelligence

Commure

Build Commure’s AI-powered healthcare claims infrastructure, including clearinghouse integrations and EDI pipelines. Develop eligibility routing and scalable revenue cycle rules.

Deschide
OpenAI

Software Engineer, Identity Infrastructure

OpenAI

Build and own secure, high-scale access proxies and identity services for OpenAI’s research and deployment platform. Take distributed systems from design through production.

Deschide