evidoria

← Back to browse

Good practice Imported

Signvrse (Terp 360) — AI Sign-Language Avatar Piloting Classroom Use from Nairobi

Kenya · Nairobi · See the Kenya profile · See the Nairobi profile

Evidence: Descriptive / self-reported Top 81% 33/100 · Ask Evidence Copilot about this practice

Kenyan startup Signvrse built Terp 360, an AI sign-language avatar, from a 20,000-sign dataset co-created with 30+ deaf signers; tested with 1,500+ deaf people in Kenya/Uganda. Pre-revenue, English-only, with acknowledged errors.

30+ signers
Deaf signers co-creating the dataset (since 2023)
20,000+ sign sequences
Captured sign sequences in dataset (since 2023)
35 %
Processing-time reduction from pre-computed sign combinations (n/a)
1,500+ deaf individuals (Kenya, Uganda)
People who tested the prototype (through mid-2025)
~2,000 users
Terp 360 web app users (as of mid-2025)
10,000 each USD (UNICEF; We Are Family Foundation)
Seed grants (n/a)
Signvrse (Terp 360) — AI Sign-Language Avatar Piloting Classroom Use from Nairobi

Details

Maturity
Pilot
Promoter
Signvrse
Period
2023–present
Keywords
assistive technology, deaf education, startup, generative AI

Context

Kenya-based Signvrse, founded in 2023 by Elly Savatia, builds an AI sign-language avatar system called Terp 360, addressing a scarcity of sign-language interpreters in East African schools.

Objectives

Create an accurate, community-grounded real-time sign-language translation tool for classroom and everyday use, starting with English-to-Kenyan-Sign-Language translation.

Activities

The dataset was built by directly motion-capturing more than 30 deaf signers to create over 20,000 professionally captured sign sequences, rather than relying on scraped or synthetic data; the company says pre-computing common sign combinations cut processing time by 35%. The prototype has been tested directly with more than 1,500 deaf individuals across Kenya and Uganda, and the Terp 360 web app has around 2,000 users.

Results

Signvrse has attracted institutional recognition: a spot among 21 organisations worldwide in Google.org's 2024 Generative AI Accelerator, Kenya's Presidential Innovation Award, the Commonwealth Secretary-General's Innovation Award, and support from Carnegie Mellon University, built on initial grants of $10,000 each from UNICEF and the We Are Family Foundation.

Conclusions

As of mid-2025, Terp 360's e-learning and classroom integrations were still described as being 'in queue', the company had not yet generated revenue, the system translated only English into Kenyan Sign Language (no Swahili support yet), and the company itself acknowledged that translation errors remained.

Implementation
Indicative cost
Low (< €50k) — Stated seed funding of $10,000 each from UNICEF and the We Are Family Foundation; low band reflects this stated minimum, actual total funding may be higher but is not disclosed.
Time to results
Medium (1–3 years) — Company founded 2023; prototype testing and institutional recognition through mid-2025 (ongoing, pre-revenue).
Staffing & skills
Signvrse founder Elly Savatia and Nairobi-based team, 30+ deaf signers who co-created the motion-capture dataset

Conditions for success

  • Community co-creation with deaf signers via direct motion capture rather than scraped or synthetic data
  • Institutional validation through Google.org, Carnegie Mellon University and national/Commonwealth innovation awards

Common failure modes

  • Pre-revenue with no confirmed business model as of mid-2025
  • English-only translation (no Swahili support yet)
  • Company itself acknowledges remaining translation errors
  • Classroom/e-learning integration still 'in queue', not deployed

Where it fits

Governance type
private startup
Scale
prototype tested across two countries (Kenya, Uganda)
Income level
low-income

Commonly funded by

Philanthropic / foundation funding

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful