ML Engineer – Generative Video
ML engineer at a young San Francisco AI lab who owns the generative video/image models behind a fast-growing creative desktop app: training and testing compact in-house diffusion/flow-matching models, running benchmarks, and turning user behavior into high-quality training data. On-site role, 3-6 years of production ML experience.
- Compensation
- $200K–$350K / yr
- Location
- San Francisco, California, United States
- Work setup
- Not specified
- Seniority
- Not specified
- Employment
- Full time
- Experience
- 3+ years
- Category
- Ml / AI
- Posted
Skills
Must have
- Personally trained or fine-tuned diffusion, flow-matching or similar generative models for images or video (API usage doesn't count)
- Taken models all the way into production; builder role, not for pure researchers or evaluation specialists
- Strong track record on the video, vision or generative-media side at a major tech company or leading generative-media lab
- Strong feel for which data improves a model, and how to gather and clean it
- On-site in San Francisco every weekday
- Roughly 3-6 years of experience
Nice to have
- A specific image or video model or feature with your name on it
- Benchmarks, evaluation tooling or data pipelines built for generative models
- Publications or open source, on top of shipped work
Full description
You've trained image or video generation models and shipped them to real users. Now own the models behind a fast-growing creative product.
A young AI lab in San Francisco makes video creation as easy as AI made coding, with a desktop app that brings a large library of AI models, search, motion graphics and editing into the AI assistant people already use. The product has strong traction, and the team is now building its own compact models to make it quicker and more capable. As their ML engineer, those models are yours: how they're trained, how they're tested, and which data makes them better.
What you'll own
- Training and testing every model that runs in the product
- Benchmarks and feedback cycles for the video models they use
- Converting real user behaviour and creative taste into high-quality training and test data
- Compact in-house models that speed the product up
Must-haves
- You've personally trained or fine-tuned diffusion, flow-matching or similar generative models for images or video (using them via an API doesn't count)
- You've taken models all the way into production. This is a builder role, not one for pure researchers or evaluation specialists.
- A strong track record on the video, vision or generative-media side at a major tech company or a leading generative-media lab
- A strong feel for which data actually improves a model, and how to gather and clean it
- On-site in San Francisco every weekday; happy to help you relocate
- Roughly 3–6 years of experience
Bonus points
- A specific image or video model or feature with your name on it
- Benchmarks, evaluation tooling or data pipelines you built for generative models
- Publications or open source, on top of what you shipped
What's in it for you
- $200k–$350k base + 0.1–0.7% equity
- Full ownership of the proprietary models on a team of fewer than ten
- Visa sponsorship available
Good to know
- Full-time, on-site in San Francisco, 9-9-5
- Process: first call with the CTO → a day working together on site → culture conversation with the CEO → references → offer
When you apply, please include
- Your CV (PDF), your LinkedIn URL, and GitHub, papers or a demo if you have them
- The model or feature you trained and shipped, and your own part in it
- Your current location, and confirmation that on-site San Francisco works for you
- Your visa or citizenship status, and if you'll need sponsorship
- Your base salary expectation, your notice period, and any other interview processes you're in