Hybrid • Paris • Full-time
At Finegrain, we believe the Internet deserves better images, and we're building the ultimate GenAI platform to make it happen at massive scale.
Our founders are repeat entrepreneurs who sold their first AI startup to Google, and we are backed by a stellar and worldwide team of investors.
Finegrain is a vertically integrated product:
- At the machine learning level, Finegrain relies on a proprietary micro framework for foundation model adaptation called Refiners. We're building it in the open (MIT license), on top of our beloved PyTorch.
- At the infrastructure level, we're building a next generation CDN to handle serving at scale of personalized images to users.
- At the application level, we provide customers with a web app that makes it simple for them to create image variations and measure user engagement, without the need for ML expertise.
Join us as a Machine Learning Research Engineer to help extend the capabilities of the Refiners framework, and train breakthrough adapters with it.
We're looking for folks who:
1. love PyTorch
2. know visual foundation models like Stable Diffusion, SAM, LLaVA inside out
3. enjoy keeping track of the latest innovations on arXiv
Why join us ?
1. You'll be part of our founding team.
2. You'll work at Station F, the world's largest startup campus.
3. You'll rub shoulders with some of the sharpest minds in AI.
4. You'll help shape a product set to break new ground.
5. You'll work on real, impactful AI. No fluff.
It is based on our bounty program: just pick one, and show us what you got. Upon completion, you'll get compensated so that you don't waste your time even if we end not working together. We typically require the completion of 2 bounties: one focused on software engineering , another one focused on training.
We feel it's the best way for you to understand what we work on and how we approach problems. Of course, you're not alone. You'll get to collaborate with the team as if you were already a part of it, through discussions on our Discord bot, and via in-depth PR reviews on GitHub.