✳MALAWI-BASED · GLOBALLY RELEVANT001/007

AI and data science solutions for local needs.

ZaZo AI helps organizations turn data and advances in artificial intelligence into useful applications. We build LLM-based tools, develop Chichewa and English customer support chatbots, and create language resources for AI research and development.

Discuss your project Explore our work
Where advanced technology
meets lived language.
FIG. 01 / LANGUAGE INTERFACE
LANGUAGE / 01 EXAMPLE
LOCAL LANGUAGE INTERFACE
✳
SAMPLE CUSTOMER SUPPORT GREETING

Moni! Tingakuthandizeni bwanji lero?

Hello! How can we help you today?
VOICE & LANGUAGECHICHEWA · NY
0.00sILLUSTRATIVE VISUALIZATION0.30s
FROM THE DATA, TO THE MODEL,
TO THE PEOPLE IT SERVES.
RESEARCH-LED. APPLICATION-FOCUSED.DISCOVER WHAT WE DO
LANGUAGE AI ✳DATA SCIENCE ✳LLM APPLICATIONS ✳SPEECH TECHNOLOGY ✳LOCAL CONTEXT ✳
01 / THE APPROACH

AI should make sense
where it's used.

Good solutions begin with the people, language and context they are built for. We combine technical depth with the practical work of preparing data, evaluating models and delivering tools that organizations can actually use.

BUILT WITH
INTENTION
02 OUR EXPERTISE

Complex technology.
Clear possibilities.

We develop AI and data science solutions that respond to real organizational needs. Explore our full range of services.

SEVEN CONNECTED CAPABILITIESDiscover all services
03 FEATURED WORK

Grounded in language.
Built through research.

From data collection to speech recognition, our portfolio examines what language AI can do for communities currently underserved by technology.

01 / SPEECH DATAOngoing

500-hour Chichewa speech collection

An ongoing Chichewa speech collection effort associated with Masakhane targeting 500 hours of transcribed audio. The work covers recording protocols, speaker metadata, transcription guidance and quality assurance to address the shortage of documented speech resources.

View project details
ZAZO / RESEARCH NOTES
500HOURS TARGET
FIG. 01 / RESEARCH & DEVELOPMENT
02 / SPEECH RECOGNITIONOngoing research

Chichewa speech-to-text research

Work dating from 2022 combines speech dataset preparation, transcript normalization, model adaptation, training and evaluation, including fine-tuning Whisper large-v3. Earlier public models and code are credited to Dunstan Matekenya.

View project details
ZAZO / RESEARCH NOTES
2022WORK BEGAN
FIG. 02 / RESEARCH & DEVELOPMENT

See ongoing projects and prototypes.

All projects
04 OPEN RESEARCH

Useful resources.
Credited correctly.

Explore Chichewa language datasets and model releases, with links to original publishers and documentation.

Dataset / Public · 2022R.01

Chichewa speech dataset

Recorded speech with transcripts, speaker metadata and dataset documentation for research and development.

Dunstan Matekenya · Zenodo
ChichewaView dataset record
Dataset / PublicR.02

English–Chichewa translation dataset

Parallel text pairs with topic and source fields for machine translation research.

dmatekenya · Hugging Face
English · ChichewaExplore dataset
PUBLIC DATASETS & MODELSVisit the resource library
THE NEXT CONVERSATION

Your next idea deserves a practical starting point.

From a question to a prototype, we can help define the right approach for the people and systems you serve.

Discuss your project