Engage Kuma
Specialist research and evaluation services for frontier labs and voice-product teams. Four engagement tiers, prices published, founder-led delivery.
| Engagement | What you get | Investment | Timeline |
|---|
| Benchmark Subscription | Quarterly evaluation of your ASR/TTS/LLM on rotating Wolof + French + code-switched test sets, with comparative analysis vs commercial alternatives. | $8,000 / month | Ongoing |
| Evaluation Engagement | One-off, deeper evaluation: failure-mode analysis, dërëm/code-switching/NER stress tests, production-readiness verdict, written report. | $15,000 – $50,000 | 4–6 weeks |
| Custom Dataset | Domain-specific Wolof or Bambara voice corpus: collection, transcription, validation, consent framework. Sample sizes 1k–50k utterances. | $30,000 – $150,000 | 6–12 weeks |
| Productized Licenses | Commercial license to wolof-numbers Pro, wolof-ner, domain dictionaries, or voice-agent templates. | $5,000 – $50,000 | Same-week |
Benchmark Subscription
Quarterly evaluation of your ASR/TTS/LLM on rotating Wolof + French + code-switched test sets, with comparative analysis vs commercial alternatives.
Evaluation Engagement
One-off, deeper evaluation: failure-mode analysis, dërëm/code-switching/NER stress tests, production-readiness verdict, written report.
Investment
$15,000 – $50,000
Custom Dataset
Domain-specific Wolof or Bambara voice corpus: collection, transcription, validation, consent framework. Sample sizes 1k–50k utterances.
Investment
$30,000 – $150,000
Productized Licenses
Commercial license to wolof-numbers Pro, wolof-ner, domain dictionaries, or voice-agent templates.
Investment
$5,000 – $50,000
Custom engagements
Production integration of wolof-numbers, wolof-ner, code-switching handlers, and domain entity dictionaries into operator voice stacks is engaged case-by-case, typically following an evaluation. Provider-agnostic; we work alongside Whisper, Gemini, Chirp, AssemblyAI, Deepgram, or your own ASR.
For engagements above $150,000 or with custom scope, contact directly.
Not sure which tier fits? Email Abdoul directly with a one-paragraph description of your problem. We reply within 48 hours with a recommendation.
Email Abdoul →If you're not yet ready for an engagement: wolof-numbers (Apache-2.0 on PyPI), wolof-ner (source-available), and the benchmark harness (Apache-2.0 on GitHub) are free to use, fork, and run on your own data.