Build local models and runtimes.
Fermion Research develops ternary QAT methods, model formats, evaluation systems, and the local runtime that serves Neutrino models. Teams are small and own complete systems.
Work arrangement
Remote-friendly, with team overlap agreed by role.
Employment
Full-time technical roles.
Application
Send relevant work, code, papers, or shipped systems.
Interview focus
Technical judgment, prior work, and collaboration.
Open roles
Full time · Remote-friendly · Systems
Member of Technical Staff, Inference Systems
Optimize single-stream decode across CUDA, Metal, ARM, and x86 while preserving output identity across backends.
Full time · Remote-friendly · Research engineering
Member of Technical Staff, Training
Develop QAT methods that let projection weights adapt to the representation used by the shipping runtime.
Full time · Remote-friendly · Evaluation
Member of Technical Staff, Evaluation
Build the infrastructure behind every published number: the accuracy batteries, the decode-rate rigs, and the token-for-token verification that gates each release.
Work spans training, model formats, evaluation, and inference runtimes.
Apply
Send what you have built to careers@fermionresearch.com
Include a short note about the work you want to do and links to the most relevant things you have built, published, or shipped.