About the role
About Sunset
Sunset is building the data layer for real-world AI training. We work with frontier labs to turn messy, multi-modal enterprise data into the highest-quality training data on the market — sourced from the hundreds of venture-backed startups we help wind down.
We're a fast-growing team based in-person in Dumbo, Brooklyn. Backed by Floodgate, Afore Capital, Hustle Fund, and incredible entrepreneurs.
The Role
As an ML Engineer at Sunset, you'll take the cleaned, resolved data coming out of our pipeline and figure out what to build with it. The raw material is unique — real codebases, tickets, messages, docs, and decisions from real companies, with every linkage preserved. The open question is how to turn that into the most valuable training data on the market. You'll have wide latitude and direct access to the CEO and CTO on direction. All of this happens on deeply sensitive data, so everything we build is designed with security and privacy at the core.
What You'll Work On
You'll own problems end-to-end. Some examples of what you might tackle in your first 90 days:
- Extracting realistic, verifiable agent tasks from linked repos, tickets, and PRs
- Building environments from real company snapshots where the reward signal comes from how work actually got done
- Augmenting datasets with synthetic variants without losing the realism that makes them valuable
- Running experiments to understand which enrichments actually move the needle for the labs buying from us
You Might Be a Fit If
- You've trained or fine-tuned models and shipped applied ML work
- You're creative and high-agency — you'd rather pick the right problem than be handed one
- You're excited about applied work with real data
