What I’m working on now
- Speech
Dialect-aware speech evaluation
Identifying when ASR systems distort or normalize real linguistic variation, and telling that apart from genuine recognition failure.
- Multilingual
Multilingual and code-switched speech
How switching between languages mid-utterance affects recognition performance, and why standard metrics misread it.
- Practice
Responsible AI in practice
Connecting fairness and human-centered design research to how AI is actually governed, evaluated, and deployed.