Machine Learning Engineer, Model Integrations

Nunchux AI; San Francisco (CA); Hybrid; USD 170,000 - 240,000; Full time

Benefits offered by this job

Health insurance

401(k)

Job summary

Nunchux AI in San Francisco is seeking an engineer to bring new image and video models to Modelverse. Your focus is speed to launch: get the model running, connect it to the platform, and ship a working first version.

You will own the initial integration, work with the cloud team on deployment and with the product engineers on the API and Modelverse integration. After launch, you will hand off further work on performance, cost, and model quality. 4 days in office, 1 day remote.

Qualifications

• Experience running and adapting generation models or integrating provider APIs.
• Ability to read model code, adapt inference pipelines, and debug model behavior.
• Familiarity with deploying models and building API integrations.

Responsibilities

• Prototype new models: run models quickly and identify launch needs.
• Ship the first version: coordinate deployment and API integration for Day 1.
• Speed up launches: automate steps, build adapters and launch scripts.
• Verify and hand off: test end-to-end integration and document limitations.
• Maintain the integration platform: fix bugs and add support between launches.

Skills

Python

PyTorch

Model integration

Production systems

Job description

About Nunchux AI

Nunchux AI builds infrastructure that makes multimodal generative AI faster and cheaper to serve, and easier to build on. Founded by MIT PhDs Muyang Li, Yujun Lin, and Zhekai Zhang with CMU Professor Jun-Yan Zhu, Nunchux brings together deep research expertise and production systems experience. Our work is built on nearly a decade of research from MIT and CMU, including nunchaku project, whose models have surpassed 4 million downloads. We have top VC backing, and we build for enterprises and for millions of developers.

The Role

Help Nunchux bring new image and video models to Modelverse as soon as they become available. Your focus is speed to launch: get the model running, connect it to the platform, and ship a working first version.

You will own the initial integration and the Day 1 release, working with the cloud team on deployment and with the product engineers on the API and Modelverse integration. After launch, you will hand off the further work on performance, cost, and model quality to the relevant teams. Between launches, you will maintain and extend our shared code for model integration and serving.

What You’ll Do

• Prototype new models: Get new image and video models running quickly with the available code and weights. Run test examples and identify what each model needs for launch.

• Ship the first version: Coordinate the deployment with the cloud team, and the API and Modelverse integration with the product engineers. Get the basic parameters, examples, and developer instructions ready for Day 1.

• Speed up the launch process: Find bottlenecks, automate manual steps, and simplify the handoffs with the cloud and product teams. Build reusable adapters and launch scripts to reduce the work each new model takes.

• Verify and hand off: Test the integration end to end, including the outputs and the error handling. Fix launch blockers, and document the known limitations for the teams that take on further optimization.

• Maintain the integration platform: Between launches, fix bugs and add support for new model interfaces, providers, and modalities in our shared integration and serving code.

What You Bring

• Image or video model experience: You have run and adapted generation models, or integrated provider APIs.

• Python and PyTorch: Strong in both, and comfortable reading model code, adapting inference pipelines, and debugging model behavior.

• Production systems experience: You have deployed a model or built an API integration. Comfortable working with existing serving tools, reading logs, and debugging failed requests.

• Working style: Quick to learn unfamiliar model code and get a prototype working. Able to keep the first release focused, resolve launch blockers, and coordinate with teammates to ship.

Bonus Points

• Experience with Hugging Face, Diffusers, ComfyUI, or model-serving frameworks such as SGLang or vLLM.

• Experience working with external model providers, or building a multi-model API platform.

• Contributions to open-source ML or inference projects.

Why Join

• Own model launches: Take new models from their first run to a release developers can use on Modelverse.

• Work with new models: Get hands-on with image and video models as they ship, across providers and architectures.

• Proven traction: Build on open-source work with more than 4 million model downloads, and on growing industry partnerships.

• The team: Work with researchers from MIT, Berkeley, and CMU, and with industry veterans from NVIDIA, AMD, Snowflake, and Adobe.

• Compensation: $170,000 to $240,000 USD base salary, plus equity and comprehensive benefits that include health insurance and a 401(k). Actual compensation will depend on relevant experience, skills, and qualifications.

Location: San Francisco, CA. 4 days in office, 1 day remote.

Start date: As soon as available

Visa: We sponsor H-1B and other work visas for exceptional candidates.

Learn more: nunchux.ai

Apply: Please apply through our Ashby careers page.

Nunchux AI is an equal opportunity employer. We celebrate diversity and are committed to creating an inclusive environment for all employees.