My mother uses a mix of Hindi and Marathi to describe how she feels. It's rarely a clinical 'abdominal discomfort'. It's more often 'thoda pet mein gadbad hai' or 'ang dhukte'. These aren't just colloquialisms - they are how health is communicated in our home. And for years, global health apps, built for an English-first world, have completely missed this.
Today marks Day 6 of our public sprint at GoDavaii, building India's Advanced Health AI. Our goal: create an intelligent assistant that speaks and understands India, in all its linguistic diversity. That means deep support for 22+ Indian languages, something no global competitor like Epocrates or drugs.com even attempts. Honestly, when we started, the sheer scale of this linguistic challenge felt daunting, but it's also our biggest differentiator.
The Unseen Language Barrier in Indian Healthcare
Think about a hurried doctor's visit in a tier-2 city. A patient, perhaps an elderly person, describes their symptoms in their native tongue - Tamil, Kannada, Bengali, Gujarati. The doctor translates, interprets, diagnoses. But what happens when families try to understand complex medication instructions, interaction warnings, or even simple home remedies from an English-only app? They're left in the dark. It's not just about word-for-word translation - it's about cultural context, nuances of expression.
This isn't just an inconvenience. It's a safety issue. My grandmother takes four different medicines every single day. For years, nobody in our family checked if those four medicines interacted. Now imagine that scenario compounded by language barriers, where even basic information is inaccessible. This is the reality for Indian families, and it's a problem GoDavaii was built to solve.
Beyond Translation: Building AI for Desi Nuances
When we talk about 22+ Indian languages, we're not just running text through a generic translation API. That's a shallow fix for a deep problem. We're training our AI Health Chat to understand the intent and context behind these varied expressions. This involves:
- Culturally-aware symptom recognition: Identifying common symptoms described in various regional idioms. A 'body ache' isn't always 'body ache' - it could be 'ang dhukte' (Marathi) or 'deham novu' (Telugu). Our AI needs to recognize all of them as the same underlying symptom, across 22 languages.
- AI-verified Desi Ilaaj (Home Remedies): This is where it gets truly unique. We cross-verify traditional Ayurvedic and home remedies with modern allopathic medicine, flagging potential interactions or contraindications. This requires a deep understanding of both systems, expressed and understood in local languages. No global player even touches this domain.
- Localizing medical information: Presenting drug interactions, side effects, and health advice in a way that resonates culturally and linguistically, making it genuinely useful for a multi-generational Indian family.
This is why building for India is fundamentally different. It's not about replicating an existing solution with a language pack. It's about designing from the ground up for a unique reality.
The Journey to Empowering 100,000 Families
We're on Day 6 of our sprint, with a public target of reaching 100,000 families. Building this kind of AI infrastructure is incredibly challenging. It demands careful data collection, robust model training, and continuous iteration. We're using state-of-the-art models like Gemini 2.5 Flash, but the real work lies in adapting them to the intricate tapestry of Indian languages and health practices.
Our aim: help families ask better-targeted questions to their doctors. We're not here to replace medical professionals, but to empower every Indian family member to be better informed advocates for their own health, in their own language.
What are some of the most challenging language or cultural nuances you've encountered in healthcare? I'd love to hear your experiences in the comments below.
-- Pururva Agarwal, Founder, GoDavaii.com
