High-Risk Pregnancy Copilot for Frontline Workers

An AI assistant helping health workers detect and manage high-risk pregnancies across India

Current Status: Deployed with 3000+ ANMs across Uttar Pradesh, Telangana and Maharashtra. 12000+ ANMs by march 2026.

Reaching: 400,000 pregnant women annually

Recognition

Winner of the Global Grand Challenges for Equitable AI 2023 by the Gates Foundation. (1500 applicants, 50 winners)

Notably, this project was featured amongst the 5 coolest innovations from across the globe by Mr. Bill Gates himself.

Winner of Google.org accelerator 2025 (3000 applicants, 20 winners)

Jump to:

• Problem

• Solution & Key Features

• Results & Impact

• Team & Partners

Problem

One woman dies in childbirth every 20 minutes in India. Over 26,000 women die annually from preventable pregnancy complications. India’s 150,000 ANMs struggle with:

• Limited access to medical knowledge during remote home visits

• Responses to questions on take 24-48 hours

• Training that fades within 30-60 days

• Overburdened with multiple digital systems

• Inadequate support for complex cases

How it works

ANMs send questions via WhatsApp (text or voice) → Copilot provides instant guidance based on validated protocols → Complex cases escalate to medical experts → Continuous learning through follow-up questions.

Key Features

• Clinically grounded: Built on ARMMAN's expert-verified protocols approved by health ministries and aligned with WHO/ICMR standards

• Culturally sensitive: Provides locally relevant advice (e.g., vegetarian nutrition for anemic patients, accessible tests vs. expensive CT scans)

• Safe & reliable: Human-in-the-loop escalation for uncertain cases; only ~20% of queries need expert review

• Fast: 8-10 second response time

• Scalable: Cost per query ~ Rs. 1.25 achieved with a mixture of product thinking and pipeline optimization

Results & Impact

Health Outcomes

ANMs report improved risk detection, better protocol adherence, more appropriate referrals, and higher quality documentation.

Adoption & Engagement

• 70% active usage in UP generating 20,000+ queries

• 80% adoption in Telangana

• 3x higher engagement vs. app-only learning

• 98% positive feedback from ANMs

Performance & utility

• 97% accuracy rated by medical experts

• 95%+ accuracy on evaluation sets

• Only ~20% escalation rate (down from 60%)

Efficiency Gains

• 60% reduction in training officer workload

• 70% faster: 8-10 seconds (from 25-30s)

• 76% cost reduction: ₹1.20 per query (from ₹5)

Team

ARMMAN: Amrita Mahale, Nithesh S, Vaishali Suyal, Dr. Aparna Hegde

Artpark: Jigar Doshi, Sid P, Sanskriti Midha, Saumitra Sharma, Nihar Desai, Raghu Dharmaraju | Past team: Apoorv Agnihotri, Mrinal Anand

Partners

ARMMAN, Gates Foundation, Google.org