Day 5 of our public sprint, and I've been spending a lot of time thinking about kaaichal. Not just 'fever' - but the way my Tamil-speaking friends describe it as a dull ache, a general unwellness that isn't quite a 'fever' in the clinical sense, but carries the same urgency. This seemingly small linguistic nuance is at the core of GoDavaii's mission and our biggest technical hurdle.
Today, we're at 379 families using GoDavaii, steadily climbing towards our public target of 100,000. Each new user brings a fresh perspective on how healthcare isn't just about medicine, but how we talk about it. As Pururva Agarwal, 27, founder of GoDavaii, I started this journey because my grandmother, speaking Hindi and Marathi, struggled to get clear health advice for her four daily medicines. This linguistic and cultural gap is magnified across India's incredible diversity.
The "Same" Symptom, 22 Different Ways
When we began building GoDavaii's AI Health Chat, the common wisdom was, "just use a good translation API." But healthcare communication isn't about direct translation; it's about interpretation. Imagine a phrase like "pet mein gud-gud" in Hindi - literally 'gurgling in the stomach'. A direct translation might miss the implications of indigestion or gas. Or a symptom like 'malaise' - how that feels and how it's described can be wildly different in Malayalam versus Bengali.
This isn't just theoretical. World Malaria Day just passed, reminding us how critical early detection and prevention are, especially for vulnerable groups like pregnant women. If a mother in a remote village describes her fatigue and chills in her local dialect, and an AI platform can't grasp the subtle, culturally-inflected meaning, we've failed her before a doctor even enters the picture. English-only apps, even the most advanced ones, simply cannot bridge this gap. This is why GoDavaii supports 22+ Indian languages - not as an afterthought, but as a foundational pillar.
Beyond Translation: The Cultural Context Layer
Our challenge extends far beyond converting words. It's about context. My grandmother might describe a remedy her mother taught her - a Desi Ilaaj (home remedy) - for a cough. A typical AI would flag it as unverified or irrelevant. GoDavaii's AI, however, cross-verifies these traditional remedies against established medical databases, bridging ancient wisdom with modern science. It's about acknowledging the reality of Indian family health practices, including fasting rituals, specific dietary habits, and traditional pregnancy care methods that are often overlooked by global health platforms.
This is where our unique moat lies. Global competitors like Epocrates or drugs.com are exceptional, but they are built for different realities. They don't understand the nuances of a multi-generational Indian home where a diagnosis for one person impacts everyone, or where language barriers often mean critical information is lost in translation during a rushed consultation.
Building the Interpreter, Not Just the Translator
Under the hood, this means we're not just fine-tuning a large language model. We're developing custom embeddings and localized knowledge graphs. We use Gemini 2.5 Flash for its multimodal capabilities and speed, but the real magic happens in the layers we've built on top. Our language models are trained on specific medical terminology and symptom descriptions across each of the 22+ languages. It's a continuous, data-intensive process of understanding not just what is said, but how it's said, and what it implies within a specific cultural framework.
For our interaction checker, which helps families like mine ensure no dangerous medicine combinations are missed, this means ingesting information that reflects local prescribing patterns and available medications - not just global standards. It's a second pair of eyes before your next appointment, helping you ask sharper questions of your doctor, augmenting their practice rather than replacing it.
GoDavaii's Sprint: 379 Families, One Big Mission
Every day of this sprint brings new learnings. We're at Day 5, with 379 users, and our target is 100,000 families across India and the world. Building in public means sharing these challenges and triumphs. This isn't just about technology; it's about trust, accessibility, and empathy.
Making AI truly useful in health means respecting and understanding local realities, not trying to force everyone into an English-only, one-size-fits-all box. It's a hard problem, but one I believe is worth solving.
What are the most challenging linguistic or cultural nuances you've encountered when building for diverse user bases, especially in healthtech? Share your thoughts in the comments.
- Pururva Agarwal, Founder, GoDavaii. Explore our work at https://www.godavaii.com
Top comments (0)