We’re an AI research lab for data. We do meaningful research, ship product, and contribute to open source. Our 4 person team has reached 7-figure ARR, with ~5x growth so far in 2025.
We train SoTA document intelligence models with novel architectures. Our tools Surya and Marker have accumulated over 42K Github stars collectively.
Our customers include tier 1 AI research labs, fast-growing startups, universities, and governments. We recently raised a seed round from founding members of OpenAI, FAIR, and Huggingface (announcing soon).
We’re looking to work with people who are low-ego, collaborative, and get things done.
We’re looking for a research engineer to work across our open source repos, inference API, and models.
You'll work at the intersection of research and engineering. You'll be training models, integrating them in our open source repos, and shipping API updates. You'll also be talking to customers and closing Github issues.
We focus on training small models with custom architectures that can outperform LLMs at specific tasks (like OCR). That said, we're not averse to fine-tuning models - we do what works best.
Day to day, you will:
Optimize inference performance
Ship features to our open source repos and API
Build clean datasets
Build and train task-specific models
Ideal Candidate
The ideal candidate for this role is someone who is passionate about building and training custom models. You’re a ruthless perfectionist who knows how to effectively navigate the tradeoffs between quality and speed.
We’re eager to work with someone who has:
Bonus points if you:
Experience building in early-stage or hyper-growth startup environments
Have publications at leading journals peer-reviewed AI conferences
Have experience building or working on OCR models
Have built an open source project or contributed to open source repositories
We believe in the simplest possible technology that gets the job done. Our stack is FastAPI (Python) for the backend and frontend, with some light HTMX and JS sprinkled in. We use Postgres and Redis, and deploy to Render. For our API, training, and inference, we mostly use Python and Pytorch.
We have a BYOD (bring-your-own-device) policy, but are also happy to offer a company sponsored laptop. You’re also welcome to choose the operating system that works best for you!
Our culture is collaborative, low-ego, and high on GSD. We aim to keep the team as small as possible, for less politics and more meaningful work.
We work out of an Industrious coworking space in Prospect Heights, Brooklyn. We value in-person collaboration, and we’re in the office most of the time.
We care about the details of how our tools are used and who is using them. It’s the big reason why we’re open source first - it enables us to have a broader impact.
30-minute video call to evaluate fit
Paid take-home project (~10 hours, $1000) - and yes, we actually do pay!
1 hour follow-up meeting to discuss the project
Culture fit interviews with the team
At this stage of company, every interview is somewhat custom, so these phases may be rearranged slightly.
Whether you’re interested in applying for one of our open roles or connecting about future opportunities, we’re eager to hear from you! Email us at [email protected] - we read every message.
P.S., We love to nerd out, so please include a link to something you’ve built or a paper you’ve published!
Compensation Range: $250K - $400K
