Skip to main content
FLAGSHIP AI CASE STUDY • ACTIVE BUILDAI & Mobile Engineering

SeHAT SmartCare AI — AI-Powered Health Assistance Platform

An intelligent Android healthcare platform that translates complex lab reports into plain language, analyzes symptoms with clinical citations, and triggers one-tap emergency GPS dispatch.

Client / OwnershipInfoHQ Internal Product
IndustryHealthcare & Clinical AI
Timeline12-Week Milestone Delivery
Core PlatformNative Android (Kotlin) + FastAPI + FAISS Vector Search
SeHAT SmartCare AI hero showcase

Key Project Results

10 WorkflowsBrought Into One Mobile AppZero app-switching friction
3 LanguagesEnglish, Hindi & HinglishWith hands-free voice input
< 1.2sVector Retrieval SpeedFAISS semantic search
100%Offline PersistenceDoze-resistant alarms & tips
System Workflow

How SeHAT Works in 3 Simple Steps

Designed so any user — from a tech-savvy student to an elderly parent — gets instant health clarity.

01

Snap a Photo or Speak Your Query

Take a clear photo of any blood test report or hold the mic button to describe your symptoms in Hindi, English, or conversational Hinglish.

02

AI Analyzes & Contextualizes

The OCR pipeline parses numerical values against standard biological intervals while FAISS semantic retrieval grounds the answer in clinical guidelines.

03

Plain-Language Advice & Action

Receive clear, understandable explanations of what the results mean, follow-up questions to ask your doctor, or set automatic medication alerts.

The Engineering Story

From Real-World Friction to Production System

Why this system had to be built, the technical hurdles encountered, and how InfoHQ solved them.

Why We Built SeHAT: Medical Reports Shouldn’t Need a Degree to Read

When families receive blood test results or pathology reports, they are confronted with dense clinical jargon, technical units, and cryptic abbreviations. Without immediate clinical context, patients experience high anxiety or turn to unverified internet forums.

Cryptic Pathology Jargon

Phrases like "Microcytic Hypochromic Anemia" or "Platelet Count 110,000 /mcL" cause immediate panic when non-clinical users cannot contextualize normal reference ranges.

The Language Barrier in India

Over 80% of Indian families think and communicate in Hindi or regional languages. Most medical apps provide English-only interfaces with zero colloquial voice support.

Fragmented Healthcare Apps

Users are forced to juggle 4 separate apps: one for lab booking, one for pill alarms, another for first-aid, and standard phone calls during critical emergencies.

A Unified, Trilingual Mobile Companion Powered by Context-Grounded AI

We engineered SeHAT as an all-in-one mobile platform combining a native Android client with high-speed Python microservices. By combining FAISS vector search with OCR and Speech Recognition, users can either photograph their reports or speak their symptoms naturally.

Instant Lab Report OCR

CameraX and PDF parsers isolate test names, values, and biological reference intervals, generating jargon-free summaries in plain English, Hindi, or Hinglish.

RAG-Grounded Medical Chatbot

FAISS vector search queries verified clinical medical guidelines, ensuring AI answers are strictly grounded and cited with standard medical disclaimers.

One-Tap Emergency GPS Broadcast

High-priority FusedLocation fixes package Google Maps coordinates and broadcast live emergency alerts to trusted family contacts via SMS and WhatsApp.

Interactive Production Explorer

Live Application Screen & Logic Breakdown

Select each workflow tab to inspect the real native interface and review how the underlying AI, OCR, and persistence pipelines operate.

Home Dashboard & Multilingual AI Medical Chatbot
Screen 01 of 05 • Native Production UI
Central Clinical Hub & Trilingual RAG Interaction

Home Dashboard & Multilingual AI Medical Chatbot

Central navigation dashboard with instant access to core health workflows, voice input integration, and an interactive RAG-powered clinical chatbot capable of responding in English, Hindi, and Hinglish with verified citations.

⚙Under The Hood • Engineering Architecture:
  • •Microphone input captured via Android SpeechRecognizer and normalized into clean text queries in English, Hindi, or conversational Hinglish.
  • •FastAPI async backend generates embeddings using SentenceTransformers and executes cosine similarity retrieval against a FAISS vector index of verified clinical literature.
  • •Context-grounded medical LLM generates answers constrained by clinical safety guardrails, citing sources and appending standard medical disclaimers.
Key Capabilities Delivered:
Direct access to clinical modulesTrilingual speech-to-text inputRAG citations & safety disclaimersInstant quick-query prompt chips
AI Symptom Checker & Medical Report Analyzer
Screen 02 of 05 • Native Production UI
Differential Clinical Assessment & Vision OCR Document Parsing

AI Symptom Checker & Medical Report Analyzer

Ranked differential symptom analysis with clinical matching (High, Moderate, Low Match), paired with a document OCR pipeline processing camera photos and PDFs of blood tests (CBC, Glucose, Platelets, Creatinine) into plain-language summaries.

⚙Under The Hood • Engineering Architecture:
  • •Differential engine evaluates multi-symptom inputs against an indexed clinical association graph to produce probabilistic tiers: High, Moderate, and Low Match.
  • •CameraX captures test documents or user uploads PDFs, passing image buffers to an OCR extraction pipeline that isolates clinical test tokens and numerical values.
  • •Parsers normalize extracted values against standard biological reference intervals (e.g. Platelets: 150,000–450,000 /mcL), generating plain-language explanations of abnormal ranges.
Key Capabilities Delivered:
Multi-symptom differential matcherCamera & PDF lab report OCRReference-interval anomaly detectionJargon-free medical explanation
Holistic Health Insights & Smart Medicine Reminders
Screen 03 of 05 • Native Production UI
0–100 Daily Health Score & High-Precision Medication Scheduling

Holistic Health Insights & Smart Medicine Reminders

Daily wellness assessment algorithm calculating a 0–100 health score with hydration and step counters, alongside a scheduled medication tracker with time-stamped dosage status and weekly adherence monitoring.

⚙Under The Hood • Engineering Architecture:
  • •Holistic health scoring algorithm computes a daily composite score (0–100) based on self-reported hydration, physical step counter integration, sleep hours, and symptom volatility.
  • •Medication alert scheduler utilizes Android AlarmManager with RTC_WAKEUP and WorkManager fallback to guarantee high-reliability alerts even under aggressive battery Doze modes.
  • •Local SQLite persistence tracks taken, snoozed, and missed doses, calculating adherence rates and syncing records with Cloud Firestore.
Key Capabilities Delivered:
Dynamic 0–100 daily wellness scoreDoze-resistant AlarmManager schedulingAdherence compliance trackingOffline-first local cache synchronization
Instant Emergency SOS & Wellness Knowledge Feed
Screen 04 of 05 • Native Production UI
One-Tap Emergency Live GPS Dispatch & Offline Preventative Care

Instant Emergency SOS & Wellness Knowledge Feed

One-tap critical SOS workflow dispatching live GPS coordinates to family contacts via SMS and WhatsApp, quick-dial 112 emergency integration, and an expert-curated lifestyle and preventative care knowledge feed.

⚙Under The Hood • Engineering Architecture:
  • •SOS workflow requests high-priority satellite & network location fixes using Android FusedLocationProviderClient, assembling an actionable Google Maps coordinates link.
  • •Direct telephony and messaging intents automate dispatch of emergency SMS and WhatsApp notifications to pre-configured trusted emergency contacts in a single user tap.
  • •Direct-dial intent triggers emergency call to 112 without intermediate confirmation screens.
  • •Preventative care articles and lifestyle advisories are pre-cached in an offline database, allowing complete offline reading during remote travel.
Key Capabilities Delivered:
One-tap live GPS emergency broadcastAutomated SMS & WhatsApp dispatchInstant 112 emergency dialer intentOffline-cached preventative health feed
Authentication & Profile Security
Screen 05 of 05 • Native Production UI
Zero-Trust Identity, Token Encryption & Isolated Clinical Profiles

Authentication & Profile Security

Secure Firebase Authentication supporting phone number verification, email/password credentials, and Google OAuth with encrypted session token handling.

⚙Under The Hood • Engineering Architecture:
  • •Multi-provider authentication powered by Firebase Auth: Phone SMS OTP, Google OAuth credential exchange, and email/password authentication.
  • •Android Keystore-backed EncryptedSharedPreferences stores JWT session tokens and user state with AES-256 GCM encryption on the client device.
  • •Cloud Firestore security rules enforce strict per-UID authorization barriers, ensuring patient medical records and reports remain inaccessible to other users.
Key Capabilities Delivered:
Phone SMS OTP & Google OAuth sign-inOn-device AES-256 token encryptionCloud Firestore zero-trust security rulesPersonalized health record isolation
System Architecture

Core System Pipelines & Infrastructure

Engineered for high concurrency, zero-trust data security, and instant response times.

01Kotlin, Java, Material 3

Android Native Client

Native Kotlin application with CameraX OCR capture, AudioRecord speech streaming, and Material 3 design tokens.

02Python 3.11, FastAPI, Uvicorn

FastAPI Microservice Backend

Asynchronous Python 3.11 endpoints handling request sanitization, token validation, and multi-threaded inference.

03FAISS, SentenceTransformers

Vector Retrieval & FAISS

Dense SentenceTransformers embeddings indexed in FAISS for sub-second semantic retrieval across clinical literature.

04Hugging Face, Medical LLM

Medical LLM & Safety Guardrails

Constrained clinical inference enforcing medical safety prompts, verified citations, and automated disclaimer injection.

05Tesseract OCR, OpenCV

Computer Vision OCR Pipeline

Image preprocessing, contrast normalization, and token extraction for blood tests, prescriptions, and lab values.

06Firebase Auth, Cloud Firestore

Cloud Sync & Device Security

AES-256 GCM token storage via Android Keystore paired with per-user isolated Cloud Firestore rules.

Full Scope Delivered

Production Capabilities & Deliverables

Every layer delivered with 100% clean code ownership, zero license hostage, and detailed documentation.

AI-assisted symptom analysis with ranked differential matches and safety disclaimers
Medical report OCR and interpretation workflow for lab tests and prescriptions
RAG-based medical chatbot grounded in verified clinical knowledge bases
Semantic medical knowledge retrieval using FAISS vector indexing
Trilingual interaction: English, Hindi, and colloquial Hinglish
Voice-based input support for hands-free queries
Cloud-backed user profile and medical report management
Medicine reminder functionality with scheduled adherence tracking
Health insights and wellness guidance with 0–100 daily score
Emergency SOS assistance workflow with one-tap live GPS dispatch via SMS & WhatsApp
Technology Stack:
AndroidKotlin/JavaPythonFastAPIMedical LLMRAGFAISSSentenceTransformersOCRFirebaseHugging Face

Have a Project in Mind That Needs Real Engineering?

From fast MVPs to enterprise AI systems and continuous retainers — let's build your next success story.