Forem: Pururva Agarwal

The Aunty Test - what Hindi-speaking patients see when they ask Health AI in their own language

Pururva Agarwal — Mon, 04 May 2026 05:01:07 +0000

Voice query in Hindi: "pet mein gas aur seene mein jalan, kya karu?"

Most Health AI built in the last three years is English-first. The translation layer breaks the moment a user types a real query in a language a billion people actually think in.

What a generic Health AI says

It seems you have stomach gas and chest burning. Please drink water and consult your doctor. Try to rephrase in English for more accurate medical information.

What native multilingual reasoning looks like

Ye acidity ke lakshan hain. 1) Thanda doodh ya nariyal pani piyo. 2) Lying down avoid karo 30 min. 3) Agar 2 ghante mein aaram nahi, antacid (Pan-D ya Gelusil) le sakte ho.

Why this is the wedge

1.4 billion people don't think in English. Every Health AI built so far has either been English-first (translate-button bolted on) or has had a thin localised veneer over an English model. Neither holds up under a real medical query.

GoDavaii reasons natively in 22 Indian languages including Hindi, Tamil, Bhojpuri, Marathi, Bengali, Telugu, Kannada, Punjabi, Malayalam, Gujarati, Odia, Urdu and more.

Free on the Play Store: https://play.google.com/store/apps/details?id=com.godavaii.app

Building publicly. Try a query in your home language and tell us where it breaks.

The Aunty Test - what Urdu-speaking patients see when they ask Health AI in their own language

Pururva Agarwal — Sun, 03 May 2026 05:00:57 +0000

Voice query in Urdu: "raat ko neend nahi aati, kya safe goli hai?"

Most Health AI built in the last three years is English-first. The translation layer breaks the moment a user types a real query in a language a billion people actually think in.

What a generic Health AI says

You mention difficulty sleeping at night. Many sleep aids exist but have side effects. Please consult a doctor and rephrase your query in English for accurate suggestions.

What native multilingual reasoning looks like

Pehle non-pharma try karein - 1 ghanta pehle screen band, room thanda (24C), magnesium glycinate (400mg). Goli zaroori ho to Melatonin (3mg) safest hai. Alprazolam / diazepam khud se mat lein - addictive hain.

Why this is the wedge

GoDavaii reasons natively in 22 Indian languages including Urdu, Tamil, Bhojpuri, Marathi, Bengali, Telugu, Kannada, Punjabi, Malayalam, Gujarati, Odia, Urdu and more.

Free on the Play Store: https://play.google.com/store/apps/details?id=com.godavaii.app

Building publicly. Try a query in your home language and tell us where it breaks.

The Aunty Test - what Marathi-speaking patients see when they ask Health AI in their own language

Pururva Agarwal — Sun, 03 May 2026 05:00:22 +0000

Voice query in Marathi: "sugar chi gholi ghetli tari BP high yetoy, kay karu?"

Most Health AI built in the last three years is English-first. The translation layer breaks the moment a user types a real query in a language a billion people actually think in.

What a generic Health AI says

It looks like your BP is high despite diabetes medication. Diabetes and hypertension are different conditions. Please rephrase in English so I can provide accurate guidance.

What native multilingual reasoning looks like

Diabetes chi golya BP la control karat nahit - donhi vegli aushad lagtat. Tumcha BP reading kay aahe? 140/90 paeksha jaast astil tar Telmisartan ki Amlodipine baddal doctor-shi boluya. Diabetic asal tar ACE inhibitor preferred.

Why this is the wedge

GoDavaii reasons natively in 22 Indian languages including Marathi, Tamil, Bhojpuri, Marathi, Bengali, Telugu, Kannada, Punjabi, Malayalam, Gujarati, Odia, Urdu and more.

Free on the Play Store: https://play.google.com/store/apps/details?id=com.godavaii.app

Building publicly. Try a query in your home language and tell us where it breaks.

The Aunty Test - what Bengali-speaking patients see when they ask Health AI in their own language

Pururva Agarwal — Sat, 02 May 2026 19:53:00 +0000

Voice query in Bengali: "matha betha ar bomi bomi bhab, ki khabo?"

Most Health AI built in the last three years is English-first. The translation layer breaks the moment a user types a real query in a language a billion people actually think in.

What a generic Health AI says

You mentioned headache and nausea sensation. Common causes include migraine. Please consult a doctor and consider rephrasing in English for clearer medical advice.

What native multilingual reasoning looks like

Matha betha + bomi bomi bhab migraine er classic combination. Andhar ghor, thanda paani-r compress, ar Domperidone (10mg) + Naproxen (250mg) khaowa jete pare. 3 din-er beshi thakle MRI kora uchit.

Why this is the wedge

GoDavaii reasons natively in 22 Indian languages including Bengali, Tamil, Bhojpuri, Marathi, Bengali, Telugu, Kannada, Punjabi, Malayalam, Gujarati, Odia, Urdu and more.

Free on the Play Store: https://play.google.com/store/apps/details?id=com.godavaii.app

Building publicly. Try a query in your home language and tell us where it breaks.

The Aunty Test - what Marathi-speaking patients see when they ask Health AI in their own language

Pururva Agarwal — Sat, 02 May 2026 18:51:11 +0000

Voice query in Marathi: "sugar chi gholi ghetli tari BP high yetoy, kay karu?"

Most Health AI built in the last three years is English-first. The translation layer breaks the moment a user types a real query in a language a billion people actually think in.

What a generic Health AI says

It looks like your BP is high despite diabetes medication. Diabetes and hypertension are different conditions. Please rephrase in English so I can provide accurate guidance.

What native multilingual reasoning looks like

Diabetes chi golya BP la control karat nahit - donhi vegli aushad lagtat. Tumcha BP reading kay aahe? 140/90 paeksha jaast astil tar Telmisartan ki Amlodipine baddal doctor-shi boluya. Diabetic asal tar ACE inhibitor preferred.

Why this is the wedge

GoDavaii reasons natively in 22 Indian languages including Marathi, Tamil, Bhojpuri, Marathi, Bengali, Telugu, Kannada, Punjabi, Malayalam, Gujarati, Odia, Urdu and more.

Free on the Play Store: https://play.google.com/store/apps/details?id=com.godavaii.app

Building publicly. Try a query in your home language and tell us where it breaks.

The Aunty Test - what Bhojpuri-speaking patients see when they ask Health AI in their own language

Pururva Agarwal — Sat, 02 May 2026 17:51:30 +0000

Voice query in Bhojpuri: "khansi do hafta se hai, ratiya mein zaada hot ba."

Most Health AI built in the last three years is English-first. The translation layer breaks the moment a user types a real query in a language a billion people actually think in.

What a generic Health AI says

It appears you have a persistent cough for two weeks worsening at night. Please see a doctor. Could you rephrase your symptoms in standard Hindi or English for more help?

What native multilingual reasoning looks like

Do hafta se khansi aur ratiya mein badhe wala matlab acid reflux ya post-nasal drip ho sakela. TB ke ruling out karaeke khatir chest X-ray + sputum test karaval ja. Honey + ginger short term raahat di.

Why this is the wedge

GoDavaii reasons natively in 22 Indian languages including Bhojpuri, Tamil, Bhojpuri, Marathi, Bengali, Telugu, Kannada, Punjabi, Malayalam, Gujarati, Odia, Urdu and more.

Free on the Play Store: https://play.google.com/store/apps/details?id=com.godavaii.app

Building publicly. Try a query in your home language and tell us where it breaks.

The Aunty Test - what Tamil-speaking patients see when they ask Health AI in their own language

Pururva Agarwal — Sat, 02 May 2026 13:46:04 +0000

Voice query in Tamil: "kaaichal moonu naal aagudhu, paracetamol thaan podhuma?"

Most Health AI built in the last three years is English-first. The translation layer breaks the moment a user types a real query in a language a billion people actually think in.

What a generic Health AI says

I notice you mentioned fever for three days. Paracetamol may help. For accurate Tamil medical guidance, please consult a Tamil-speaking physician or rephrase your query in English.

What native multilingual reasoning looks like

Moonu naal mela kaaichal irundha paracetamol madhum podhadhu. Dengue, typhoid, malaria check pannanum. CBC + dengue NS1 test edukkanum. Indha 24 mani neram-la doctor-a paarunga.

Why this is the wedge

GoDavaii reasons natively in 22 Indian languages including Tamil, Tamil, Bhojpuri, Marathi, Bengali, Telugu, Kannada, Punjabi, Malayalam, Gujarati, Odia, Urdu and more.

Free on the Play Store: https://play.google.com/store/apps/details?id=com.godavaii.app

Building publicly. Try a query in your home language and tell us where it breaks.

The Aunty Test - what Hindi-speaking patients see when they ask Health AI in their own language

Pururva Agarwal — Sat, 02 May 2026 07:02:55 +0000

Voice query in Hindi: "pet mein gas aur seene mein jalan, kya karu?"

Most Health AI built in the last three years is English-first. The translation layer breaks the moment a user types a real query in a language a billion people actually think in.

What a generic Health AI says

It seems you have stomach gas and chest burning. Please drink water and consult your doctor. Try to rephrase in English for more accurate medical information.

What native multilingual reasoning looks like

Ye acidity ke lakshan hain. 1) Thanda doodh ya nariyal pani piyo. 2) Lying down avoid karo 30 min. 3) Agar 2 ghante mein aaram nahi, antacid (Pan-D ya Gelusil) le sakte ho.

Why this is the wedge

GoDavaii reasons natively in 22 Indian languages including Hindi, Tamil, Bhojpuri, Marathi, Bengali, Telugu, Kannada, Punjabi, Malayalam, Gujarati, Odia, Urdu and more.

Free on the Play Store: https://play.google.com/store/apps/details?id=com.godavaii.app

Building publicly. Try a query in your home language and tell us where it breaks.

Day 9: Building GoDavaii - From Healthcare Gaps to 22+ Language AI Trust

Pururva Agarwal — Tue, 28 Apr 2026 16:13:34 +0000

It's Day 9 of our public sprint with GoDavaii. At 27, I founded this company with a deep conviction: the existing healthcare system, especially in India, often fails families. Not for lack of dedicated doctors, but for a critical gap in accessible, trustworthy information - particularly in the language people actually speak and think in. When I see headlines about cancer striking younger people in India, it underlines the urgency. The medical knowledge exists, but if it's trapped behind English-only interfaces or complex jargon, it creates a systemic barrier.

My personal journey into this space was forged watching my own family navigate these exact challenges. When health questions arise, especially complex ones like medication interactions or understanding a lab report, the first instinct for many isn't a doctor's appointment - it's often a frantic search, sometimes in a language Google Translate barely handles, sometimes asking a relative who's 'heard something'. This isn't just about convenience; it's about dignity and efficacy. GoDavaii is designed to bridge that gap, not by replacing doctors, but by empowering families with a thinking tool that helps them ask better-targeted questions.

The Unseen Complexity of 22+ Indian Languages

Most global health AIs are built for English-first markets. Epocrates, drugs.com, Medscape - powerful tools, but they don't speak Tamil, Hindi, Marathi, or Gujarati. This isn't just a translation problem; it's a contextual, cultural, and semantic challenge. Our AI Health Chat in 22+ Indian languages isn't just translating medical terms; it's trained to understand nuanced expressions. For instance, in Tamil, saying "shareeram sariyaagilla" (I'm not feeling a little well) isn't just a general complaint; in a health context, it's a symptom description that the AI must learn to parse effectively. This is the heart of building for the next billion: reaching people who are coming online with health questions that English-centric AI simply cannot understand.

Developing this requires a deep understanding of linguistic variations, local idioms, and even the common ways people describe symptoms in their mother tongue. We're talking about fine-tuning large language models not just on medical texts, but on real-world, colloquial health conversations across a vast linguistic landscape. It's technically challenging, and honestly, it's our biggest moat against global players.

AI-Verified Desi Ilaaj: The Crossroads of Tradition and Tech

Another unique challenge, and a core feature of GoDavaii, is AI-verified Desi Ilaaj. This isn't about promoting unproven remedies; it's about acknowledging that for generations, families in India have relied on home remedies, often passed down verbally. The problem starts when these traditional practices, many of which have genuine therapeutic value, encounter modern allopathic medicines. What happens when a commonly used Ayurvedic concoction interacts negatively with a prescription drug? No global health AI even attempts to cross-verify these combinations. They simply don't have the cultural context or the data.

Our system uses AI to cross-reference known Ayurvedic ingredients and traditional practices with a comprehensive database of allopathic drug interactions. It's about providing a layer of safety and understanding that has been entirely absent. It's a delicate balance - respecting tradition while ensuring modern medical safety. GoDavaii aims to surface potential interactions, helping families understand the risks and enabling them to have informed conversations with their doctors. We're an information layer between you and the doctor that offers diagnoses or guarantees cures.

Building in Public: The Hard Truths of Day 9

As of Day 9, we're still at 0 users. Yes, zero. Building in public means sharing the raw, unvarnished truth. We're iterating, we're refining, and we're ensuring the core product delivers on its promise of accurate, culturally relevant information. The focus right now is less on vanity metrics and more on strengthening the underlying AI models for accuracy in areas like our Drug Interaction Checker and the Pregnancy medicine safety checker. We're laying the foundation for scale, ensuring that when the next wave of users comes, the experience is robust and reliable.

My biggest lesson so far? Trust is built byte by byte, language by language. It's not about being the loudest or having the biggest numbers; it's about deeply understanding the user's unmet need and delivering a solution that genuinely helps them navigate their family's health. The AI doesn't just need to be smart; it needs to be empathetic to the user's specific cultural and linguistic reality.

What's one health question you've had that you wish an AI could answer accurately in your native language, or a traditional remedy you've wondered about combining with modern medicine? Drop your thoughts or questions in the comments below - I'm reading every one.