HACKYARD BUILD 2026
Sushruth
AI healthcare companion
Team YOLO
Anushka Singha
Sharique Ahmed
Anuska Das
Faiyazul Islam
A multilingual AI healthcare assistant that helps people understand symptoms and reports, organize health records, and bridge the gap between history-taking and consultation with clear summaries for clinicians.
Problem: Patient context gets lost between steps
50
study participants
11
languages spoken
Kozan et al., Frontiers in Communication
(2026),
doi:10.3389/fcomm.2026.1848819
Patients, caregivers and clinicians can face
language barriers and fragmented health
histories.
Participants in a Karnataka study described
added time, cost and stress associated with
language discordance.
Local qualitative evidence—not a national
estimate or a measured Sushruth outcome.
Source: Kozan et al. · Frontiers in Communication · 2026
• Health information is scattered across reports, prescriptions and patient recollection.
• Medical terminology and language barriers make information hard to understand.
• Paper records can be lost or difficult to retrieve during a consultation.
Patients often depend on paper, manual explanations and separate tools—with limited continuity between visits.
How it is being handled currently:
What is in the current Sushruth app
CORE WEBSITE FLOWS
BUILD 2026 · AGENT 37 OFFLINE ADDITION
Health-record edit
→
Local Pending
→
API confirms
→
Synced
Stack: React/Vite · Clerk · Express + Zod · PostgreSQL/Drizzle · OpenAI · Claude
Multilingual healthcare support (22 Indian languages)
AI triage, pre-consultation, OCR and health records
Care Circle and emergency access
Agent37-assisted development and offline sync
System Flow�Patient Input → AI Processing & Record Management → Structured Health Information
Technologies
React + Vite — Frontend
Clerk — Authentication
Express.js + Zod — API & validation
PostgreSQL + Drizzle — Database
Agent37 — Development assistance
Build Contributions
Existing: Core Sushruth healthcare platform
Implemented/refined: Offline health-record queue, retry and sync-status tracking
Next: Stronger offline security and further integration testing
Working demo: save offline, sync on reconnect
Edit synthetic record → save while offline → Pending → reconnect → API confirms → Synced (Built during Hackyard build)
ADD SCREENSHOT · PENDING
ADD SCREENSHOT · SYNCED
6 / 6
Focused offline-queue tests passed
Not verified: live PostgreSQL integration or clinical
validation. Local storage is unencrypted; the app
must already be loaded. ABHA/AADHAR integration not yet. Real clinic
Verification not yet developed.
A distinct visit-preparation focus
EXISTING PLATFORMS
MyChart lists provider-linked records and
a symptom checker. Ada presents an AI
symptom assessment.
SUSHRUTH CURRENT PROTOTYPE
Connects triage, report explanation and patient history for visit
preparation; BUILD adds offline record edits. Also supports
Multilingual and accessibility options for visually Impaired and
the elderly.
Difference: workflow focus—not a claim of better diagnosis, broader integrations or platform
replacement.
Technical decision: keep edits Pending until
API confirmation; retain failures for retry.
Challenge: intermittent connection; explicit
Pending/Synced states make retry outcomes
visible.
Sources: mychart.org/Features · ada.com/app
From prototype to supervised care
WHO IT MAY SERVE
Patients and caregivers preparing
health context.
Clinicians receiving a structured
intake handoff.
Potential value: clearer context
before visits; impact is not
measured.
BEFORE REAL-WORLD USE
Secure local health data and review privacy
controls.
Test retries with PostgreSQL and handle
cross-device conflicts.
Complete clinical validation and a supervised
pilot.
ABHA / ABDM / FHIR remain future scope—not
current integrations.
Build next: secure, tested handoff that can be evaluated with patients and care teams.