Dendrite -agent for community health workers

India has over a million ASHA and Anganwadi workers — frontline community health workers who visit homes across villages and urban slums to check on pregnant women, track child vaccinations, monitor newborns, and follow up on TB and malaria cases. Every one of those visits is recorded by hand in a paper register, then manually re-typed into district health databases days or weeks later. That delay is a major reason India's public health surveillance data is often stale and error-ridden by the time anyone acts on it. Dendrite replaces the paper register with speech. After a visit, the worker simply talks — in Hindi, English, or the natural mix of both that people actually speak. AssemblyAI's Universal-3.5 Pro Streaming model transcribes the audio in real time with native code-switching support, so the worker watches their words appear as they speak. An LLM then extracts the visit into structured fields: patient, symptoms, vitals, vaccination status, and follow-up date. If a required field is missing, the agent asks one short clarifying question out loud rather than silently leaving a blank, then reads back a one-line summary for confirmation before saving. The structured record reaches a supervisor dashboard within seconds, alongside the full raw transcript kept for audit. Because visits link to persistent patient records, the app also computes what is due: given a child's date of birth, it knows every upcoming and overdue vaccine from India's National Immunization Schedule and surfaces that to the worker at the start of a visit. The result is the same information a worker already collects, captured in 30 seconds of speech instead of handwriting, with zero re-typing and no weeks-long lag between what happens in the field and what health authorities can see.