We publish what we learn.
Notes, methods, and datasets on how people communicate under stress and uncertainty, and how we build agents that listen well.
-
Listening when words are hard
Why the conversation before a consultation is the hardest five minutes in a hospital, and how we design agents to read intent, not just keywords.
Read → -
IndMedSTT-Synth: when speech-to-text mishears medicines
A public dataset of how multilingual ASR garbles Indian medication brand names, with the true drug, the clinical context, and the exact misheard phrase.
View dataset →