Gradr is a Stockholm-based AI platform for Swedish schools. Teachers run both assignments and exams on Gradr's own exam platform, the AI drafts formative feedback, and the teacher reviews and approves it before a student sees anything. Gradr deliberately stays out of summative grading, and that choice is a large part of why schools trust it.
This is the role that owns the model work behind that feedback. Quality of feedback is the core of the product, so this is not a wrapper around an API. It is the difference between a teacher approving a draft and rewriting it.
What you will do
- Own the systems that draft formative feedback, and the evaluation that tells you whether the drafts are actually good
- Work on the harder half of the product: using AI to help a teacher understand where a particular student is strong and where they need help
- Build for a workflow where a human always approves before a student sees anything, which changes what a good output looks like
- Work on Swedish-language education material, where general-purpose models are weakest
What we are looking for
- Strong Python and practical experience taking language models into production, including evaluation rather than just prompting
- Judgment about output quality, and the patience to measure it rather than eyeball it
- Interest in education, and in tools where a professional stays in the loop by design
Nice to have
- Experience with evaluation harnesses, structured output, or retrieval
- Work in a regulated or high-trust domain
- Swedish, or a willingness to work closely with Swedish-language material
About the company
- Launched in August 2025 and now used by more than 100 schools
- Revenue tripled in the first half of 2026
- Customers include Nacka Gymnasium, Birger Jarls Gymnasium and Söraskolan, plus an ongoing pilot with Stockholm stad
- White-label partnership with SchoolSoft, one of the largest school platforms in Sweden, and another partnership with Unikum on the way
- A team of five, three on product and two on go-to-market