Nunchux AI is hiring a Machine Learning Engineer, Model Integrations in San Francisco, offering $170,000–$240,000 base salary plus equity. The full-time role involves integrating image and video generation models, production API development, and model deployment. Candidates need 6–9 years of experience, with H-1B and other visa sponsorship available.
Machine Learning Engineer, Model Integrations
Full-time role with Nunchux AI, based in San Francisco, California — 4 days in office, 1 day remote.
Machine learning engineer jobs with real equity and H-1B backing don’t show up every day. This role at Nunchux AI puts you at the center of every new model launch — from the first prototype run to a live release millions of developers can use — with a compensation package that actually matches the work.
About the Role
Nunchux AI builds the infrastructure that makes multimodal generative AI faster and cheaper to run, and easier for developers to build on. Three MIT PhDs — Muyang Li, Yujun Lin, and Zhekai Zhang — founded the company alongside CMU professor Jun-Yan Zhu, and their open-source nunchaku project has already crossed 4 million downloads.
This job is about speed. When a new image or video model drops, you’re the one who gets it running on Modelverse first. You prototype it, wire it into the platform, and ship a working Day 1 release — then you hand the deeper optimization work to the teams that specialize in performance and cost.
Between launches, you’ll maintain and extend the shared code that makes every integration faster than the last. Think reusable adapters, launch scripts, and fewer manual handoffs with the cloud and product teams each time a new model needs a home.
You’ll work alongside researchers from MIT, Berkeley, and CMU, plus industry veterans who came out of NVIDIA, AMD, Snowflake, and Adobe. It’s a small, technical team backing a product that’s already proven itself with millions of downloads and a growing list of enterprise partners.
Primary Responsibilities
- ✓
Prototype new image and video models quickly using available code and weights - ✓
Run test examples and flag exactly what each model needs before launch - ✓
Coordinate Day 1 deployment with the cloud team and API integration with product engineers - ✓
Build reusable adapters and launch scripts that cut the work each new model takes - ✓
Test each integration end-to-end, including outputs and error handling - ✓
Document known limitations so downstream teams can pick up optimization work - ✓
Maintain and extend the shared integration and serving code between launches
Qualifications
- ✓
6 to 9 years of experience running or adapting image and video generation models, or integrating provider APIs - ✓
Strong in Python and PyTorch, comfortable reading model code and debugging inference pipelines - ✓
Hands-on experience deploying a model or building a production API integration - ✓
Bonus: experience with Hugging Face, Diffusers, ComfyUI, SGLang, or vLLM
Job Snapshot
Full-time
Mid-Level (6–9 years)
Bachelor’s degree
$170,000–$240,000 USD base, plus equity
4 days in-office, 1 day remote
San Francisco, CA
H-1B and other work visas sponsored
As soon as available
Why Join
If you like being the first one to get a new model running, this role is built for you. Nunchux is scaling fast and needs someone who can turn a research checkpoint into a shipped product without losing momentum.
Ready to Apply?
Hit the button below to go straight to the official job listing.
Prefer to go straight to the source? You can also apply directly through Nunchux AI’s official careers page (hosted on Ashby), linked from their site at
nunchux.ai
.