How they use Nebius AI Cloud
Lity Club built a document-recognition component for its legal copilot using Nebius Serverless AI. The project runs Qwen2.5-VL-7B-Instruct through vLLM on an H100 SXM endpoint. It extracts structured identity-document fields, assigns per-field confidence scores and supports multi-page PDFs. Nebius Object Storage holds recognition blueprints and results.
The repository documents live recognition runs and an evaluation on 60 synthetic identity documents. The endpoint is deployed on demand and stopped between runs. Nebius named Team Lity the challenge’s HCLS special-award winner. Dmitry Sadchikov said award credits would support MVP work and described a future multicloud production setup including Nebius.
Sources
3 sourcesdocs-proc-nebius
Listed 20 Oct 2026 · GitHub · Web page
The project documents a Nebius Serverless GPU Endpoint running document extraction through vLLM, live execution artifacts, synthetic-document evaluation and Nebius Object Storage for blueprints and results. The endpoint runs on demand.
Open original sourceNebius Serverless AI Builders Challenge: winners announced
Published 18 Aug 2026 · Nebius · Web page
Nebius names Team Lity as its HCLS special-award winner for serverless document recognition supporting a legal copilot and links the project repository.
Open original sourceCompany context
lity.clubLity Club is an early-access AI legal analysis and lawyer-matching platform operated by UAB Redela and UAB IPLK in Vilnius, Lithuania. It helps individuals describe legal problems in everyday language, understand relevant laws and prepare case files before consulting a lawyer. Its Spanish-language service focuses on immigration and housing rental, with explanations in Spanish, English and Russian. The platform also targets lawyers with prepared client cases and AI workflow tools; broader matching, payments and collaboration features remain under development.
Lity Club official website and Terms of Service