Sepal is a data research company with a mission to advance human knowledge & capabilities through safe AI. We partner with researchers at the frontier AI labs & leading enterprises building models.
Right now, we’re primarily focused on building RL environments and evals:
- Hundreds of our SWEs and experts are currently building environments and outcome-verifiable tasks for our customers.
- Our focus (so far) has been in coding, finance, and enterprise knowledge worker environments. We’re actively adding to this.
What we’ve built to date:
- A framework for environments and tasks and a platform + operational process for building RL environments & tasks.
- An expert network (originally to power traditional annotation and testing) which were now mainly leveraging to design new real-world environments and tasks.
- Previously we’ve build some of the strongest internal reasoning benchmarks (sub 5% pass rates on frontier models like o3) for the world’s leading AI labs - quoted by researchers as ‘the most powerful benchmark our company has access to’.
Our team comes from Turing, Amazon, Vercel, McKinsey, Bain, Stanford, quant trading, & more.
---> Now about you.
- If you’re a researcher and are interested in accessing some of the environments we’ve built, reach out at hello[at]sepalai[dot]com.
- If you’re a software engineer, AI researcher, or growth operator and want to join our company, click here.
- If you’re a leading expert in your field and interested in joining our network to get paid in designing these environments, click here.