Featured by CHIA: Helping Quechua Communities Shape Their Future with AI

The Centre for Human-Inspired Artificial Intelligence (CHIA) at the University of Cambridge has featured my work in its Meet the Researcher series:

Meet Elwin Huaman: Helping Quechua Communities Shape Their Future with AI

The feature tells the story behind QuechuaBase and our work with Quechua communities in Puno, Peru. It covers:

• Where it started. I grew up in Nuñoa, at 4,100 metres in southern Peru, speaking Quechua, a language that lives mostly in speech and has had little support in digital spaces. You can read more in my story.
• Community-led data collection. In January and February 2026, together with the National University of Altiplano of Puno, the Puno Regional Government and the Illariy Ch’aska association, 151 volunteers contributed 66 hours of Puno Quechua speech. Community members also helped decide what to record and what counts as non-sensitive. The details are in our AmericasNLP 2026 paper.
• Speech first. Many speakers are not literate in Quechua, so technology that depends on reading and writing leaves them out. A speech recognition model of around 317 million parameters, adapted on community data, outperformed a 7.8-billion-parameter multilingual model on in-domain tests. That result does not hold in every setting, and the feature argues for evaluation that also looks at real-world use, device access, cultural fit and performance across groups of speakers.
• What comes next. Text-to-speech voices, emotion recognition for Quechua, and a shared task to grow language technology for Indigenous languages across South America.

One line from the interview sums up how I see this work:

“Communities should hold their own data and be the ones who validate what machines produce.”

Thank you to CHIA for the feature, and to the Quechua speakers, partners and volunteers in Puno who made this work possible.