I am Asha

Your Key to Wellbeing

The fiduciary medical AI in the DNAi Healthcare Fleet. Ask any health question and receive an evidence-grounded answer with transparent citations from clinical references, drug labels, and peer-reviewed studies.

Ask Asha Now See How It Works

Built by physicians · Patient-facing · Multilingual via Gemini 3 · Free tier

5.1M
PubMed Abstracts
889K
DailyMed Drug Labels
76K
StatPearls Articles
11
Languages

What Asha Does

Asha is the patient-facing intelligence in the fleet. She covers the questions a thoughtful patient or caregiver actually brings to a clinic visit.

📝

Plain-language health answers

Symptoms, conditions, lab results, side effects. Asha explains the medicine in accessible language without dumbing it down.

💊

Medication safety

Drug interactions, common side effects, contraindications, pregnancy and lactation safety. Grounded in DailyMed and FDA labels.

📖

Evidence with citations

Every answer carries clickable references back to the original literature. Asha never fabricates a citation; if no evidence exists, she says so.

🌍

Multilingual care

Ask in English, Hindi, Spanish, Chinese, Arabic, German, French, Portuguese, Russian, Japanese, or Korean.

🎙

Voice in, voice out

Hands-free conversations for the kitchen, the car, and the caregiver at the bedside.

🚨

Emergency escalation

When a question signals an emergency, Asha breaks the conversation to point you to 911 or local emergency services. This behavior is architectural.

How Asha Is Grounded

Asha pulls evidence from the DNAi corpus before answering. The highest-weighted collections appear first in her retrieval, with boost values defined in agents/asha/agent.yaml.

Collection Boost Scale What it carries
statpearls_clinical 1.5x 76K articles Peer-reviewed clinical reference, the bedrock of bedside teaching
dailymed_drug_labels 1.5x 889K labels Full FDA-approved drug labels, dosing, indications, warnings
clinical_guidelines 1.4x varies Specialty-society guidelines (cardiology, oncology, pediatrics, more)
pubmed_abstracts 1.4x 5.1M abstracts Peer-reviewed biomedical literature
healthbench_rubrics 1.3x varies Quality rubrics used by the OpenAI HealthBench evaluation
Mode all: beyond these boosted collections, Asha has unrestricted retrieval across the full corpus of roughly 829 live collections and over 126 million vectors. Nutrition, rare disease, mental health, pediatrics, geriatrics, humanitarian medicine, and traditional pharmacopoeias are all in scope. She does not say "my sources do not contain that" for a question another fleet agent could answer.

Sacred Refusals

Twelve constraints are encoded into the architecture, not bolted on as policy. Numbers below match the canonical list in core/prompts.py.

1

Never prescribe

Asha will not write prescriptions or recommend specific doses for a named patient. Drug information is reference-grade, not order-entry.

2

Never diagnose definitively

Differential reasoning, yes. Decisive diagnostic labels, no. That belongs to a clinician with the patient in front of them.

3

Never browse live URLs

Asha cannot fetch the live web at answer time. Every claim is grounded in the indexed corpus or hedged accordingly.

4

Never override emergency escalation

Suicidality, chest pain, anaphylaxis, stroke symptoms: Asha pivots to emergency guidance and will not be talked out of it.

5

Never claim absolute certainty

Medicine is probabilistic. Asha reports confidence honestly and flags when evidence is thin or contradictory.

6

Never abandon compassion

Efficiency does not outrank care. A frightened parent at 2 a.m. gets warmth as well as evidence.

7

Never prioritize fluency over safety

A smooth answer that buries uncertainty is a fiduciary failure. Asha would rather sound careful than sound clever.

8

Never proceed without evidence

If the question requires evidence and none is retrieved, Asha says so and refuses to confabulate. Silence is a valid clinical answer.

9

Never fabricate PMIDs

If a citation cannot be retrieved from the corpus, it does not appear in the answer. Made-up references are unreachable.

10

Never cite outside the retrieved set

Asha will not invoke references that were not surfaced by retrieval. The provenance chain stays unbroken.

11

Never reassure during active symptoms

When the user is still experiencing symptoms in the conversation, future-tense reassurance is prohibited. Asha stays present with what is happening now.

12

Minor in crisis: request a parent

If a minor is in active crisis, Asha will not continue clinical content alone. A parent or guardian must be brought into the conversation.

Languages

Eleven languages are first-class, with linguistic and cultural grounding rather than literal translation.

South Asia and Middle East

English, Hindi (हिंदी), Arabic (العربية). Asha pulls from regional clinical and nutrition collections, including the Indian National Database for foods.

Europe and the Americas

Spanish, Portuguese, French, German, Russian. Useful for diaspora caregivers and travelers.

East Asia

Mandarin Chinese, Japanese, Korean. Voice support included.

Pricing

Asha is free to try with a generous quota. Pro unlocks unlimited queries, document upload, and a personal health memory you control.

Free tier: includes evidence citations, voice, and the full medical corpus. Pro: unlocks document analysis, longitudinal memory, and unlimited queries. See the full pricing page.

Frequently Asked Questions

Is Asha a doctor?
No. Asha is an evidence-based medical AI built by physicians. She provides educational health information and is not a substitute for professional medical advice, diagnosis, or treatment.
How is Asha different from a general chatbot?
Every response is grounded in retrieved evidence from a dedicated medical corpus of around 829 collections totaling over 126 million vectors. Citations are clickable and verifiable. If the corpus has no answer, Asha says so rather than guessing.
What happens during an emergency?
Asha detects emergency signals (suicidality, chest pain, stroke symptoms, anaphylaxis, severe pediatric distress) and pivots to local emergency-services guidance. This behavior cannot be turned off in conversation.
Is my data private?
Yes. Data in transit uses TLS 1.3, personal health information is anonymized, and your conversations are never sold. You can delete your data at any time.
Does Asha cover nutrition and mental health?
Yes. Asha inherits the nutrition corpus from her sibling Sage and a dedicated psychiatry stack including assessment scales, seminal trials, and consultation-liaison material. She covers the full biomedical scope a patient might ask about.
Who built Asha?
DNAi Systems, founded by Deepan Singh, MD, FAPA (psychiatrist) and Paridhi Anand, MD (pediatrician). The platform runs on a patent-pending fiduciary cognitive architecture.

Ready to ask Asha?

Evidence-grounded answers to your health questions. Free to start. No card required.

Try Asha Free