Terp 360 — Signvrse's AI Sign-Language Avatar for Kenya's Deaf Community

Kenya has an estimated 2.7 million Deaf and hard-of-hearing people, yet African countries have as few as 40–700 certified sign-language interpreters between them, leaving Deaf citizens locked out of healthcare, education, banking and government services. Founded in Nairobi in 2023 by engineer Elly Savatia, Signvrse set out to close that gap with software instead of scarce human interpreters.
Terp 360 uses motion-capture recordings of professional Kenyan Sign Language (KSL) signers, layered with a generative-AI translation model, to render spoken or typed English as a photorealistic 3D avatar signing in real time. The training dataset — more than 20,000 professionally captured sign sequences, drawn from signers in both urban and rural Kenya — was built and validated in direct collaboration with Deaf community members, sign linguists and AI engineers, and pre-computed sign combinations cut processing time by roughly 35%.
Early prototypes, built on basic Unity animation, produced stiff, robotic motion that Deaf testers found hard to read; the team rebuilt the animation pipeline after that feedback to produce more natural, fluid signing — a concrete example of iterating with the community it serves rather than for it. As of its 2025 beta, the web-based platform had engaged more than 2,000 Deaf and hard-of-hearing users for testing across Kenya and Uganda, and the company has opened a second office in Rwanda to pilot classroom and public-service deployments.
Signvrse has been backed by $10,000 grants each from UNICEF and the We Are Family Foundation, was one of 21 organisations worldwide selected for Google.org's 2024 Generative AI Accelerator, and won Kenya's Presidential Innovation Award and the Commonwealth Secretary-General's Innovation Award, besides being a Microsoft Imagine Cup finalist; founder Elly Savatia was named to the 2025 Africa Prize for Engineering Innovation cohort. The company plans to move to a B2B SaaS licensing model, with initial revenue targeted from the edtech sector, from the fourth quarter of 2025.
Independent, peer-reviewed evaluation of translation accuracy has not yet been published, and Terp 360 remains one-directional (spoken/written language to sign, not sign to speech), so it does not yet replace two-way human interpretation.

Read the full analysis: https://www.signvrse.com