/ careers
Research Engineer, Training & Data Systems
Remote / San FranciscoFull-time$150K - $250K0.5% - 2% equity
The Role
You'll work with the founding team to study where models fail on document-centric tasks, turn those failures into training problems, and build the datasets and environments needed to fix them. This is a small team where decisions happen fast and there's no layer between you and the problem.
What you'll do
- Dataset curation, sampling, shape, and the feedback loops between data quality and model quality
- Train and improve models across architectures (object detection, VLMs, LLMs) using supervised and RL approaches
- Build evaluation pipelines and error analysis workflows that guide what the team works on next
What we're looking for
- Experience with RL post-training, reward modeling, or the data pipelines that feed them
- You've trained at least two of: object detection models, VLMs, diffusion models, LLMs
- Experience with document AI or agent workflows on real-world use cases
- Solid research taste. You know which ideas are worth trying and you're right more often than not
- Clear technical writing
What would impress us
- Experience at scale: large datasets, distributed training, or meaningful GPU/compute usage
- A corpus of technical work we can look at: OSS PRs, Hugging Face models, papers, technical blogs, or detailed public repos
Benefits
- Premium medical, dental, and vision
- Daily meal stipend
- Monthly travel stipend
- Flexible time off
- Relocation stipend
Tech stack
PythonTypeScriptPostgreSQLNext.jsBun/UVGCP/AWSKubernetesDockerBare-metal GPU providersCloudflare
Unlimited token budget.
How to apply
Email careers@floatingpoint.ai with a resume and a few sentences on what interests you about the work. We like seeing personal projects, writing, and open-source contributions.
