Re-skill

Re-skill builds AI software for enterprise training, turning internal policies, manuals and wikis into curricula, lessons, videos, voiceovers, quizzes and slides. Training teams can connect Notion, Confluence or document stores to create onboarding, compliance and product training. Subject-matter experts review each screen before release. Courses export through SCORM or xAPI into customers’ existing learning management systems, keeping enrolment, reporting and administration in place.

Founded by Farrukh Umarov and Said Homidov—identified in 2025 as CEO and CTO, respectively—Re-skill originated as a spin-off of the skills accelerator ilmhona. It initially served students and individual learners before shifting to corporate training in 2025. Its engineering work also includes a Tajik text-to-speech checkpoint trained on 35 hours of audio. Adapted from Orpheus, it combines a Llama-3B backbone with SNAC, a neural codec that converts audio into discrete tokens for next-token prediction.

Re-skill prices plans by course volume, with cloud delivery by default and VPC or on-premise deployment on Enterprise plans. Its customer case study attributes to Arvand Bank’s learning and development team a 2025 estimate of approximately 20 hours of manual work saved weekly, alongside training-completion visibility across roughly 1,000 employees.