Healthcare AI Platform — India

MSR Pulse
Integrated Clinical AI

6 specialized AI microservices unified by an LLM-powered agentic orchestrator — from differential diagnosis and ED patient flow forecasting to treatment pathways, clinical note search, and hospital resource optimization for M.S. Ramaiah hospitals.

🏗️ System Architecture 🧠 Explore AI Modules ⭐ Key Advantages
MSR Pulse AI Healthcare Platform — Clinical AI Dashboard
0AI Services
0Containerized Services
0Agentic Tools
0Lines of Code
0Patient Records
0Clinical Notes
System Architecture

6 AI Services + 8 Infrastructure Services

A clinician asks a natural-language question through a single conversational interface, and the system automatically classifies, routes, calls downstream services, synthesizes results, and flags safety-critical decisions.

MSR Pulse System Architecture — 6 AI Microservices

🧠 MSR Pulse Agent

LangGraph + GPT-4o Orchestrator — Port :8100

Router (GPT-4o-mini) Reasoner (GPT-4o) Validator (Approval Gate)
🩺

PulseDiagAgent

Differential Diagnosis via GPT-4o CoT

:8005
📈

PulseFlow

ED Patient Flow Forecasting via LSTM

:8001
💊

CarePlanPlus

Treatment Pathways via BERT

:8002
📋

PulseNotes

Clinical NLP & RAG Search

:8003
🏥

MediSync

Multi-Agent RL Simulation

:8004
💬

Dashboard

React + TypeScript + Vite

:5173
MongoDB 6.0PrometheusGrafanaMLflow v2.9MinIORedis 7NginxDocker

⚡ Intelligent Query Flow

User Query → Router (GPT-4o-mini ~200ms) → Reasoner (GPT-4o) → Tool Executor → Validator ⚠️ → Clinical Response

Dual-LLM Architecture: GPT-4o-mini for routing (10× cheaper) + GPT-4o for reasoning — reducing API costs by ~60%

AI Modules

6 Specialized AI Microservices

Each service can be scaled, updated, or replaced independently — microservice independence with agentic orchestration.

🩺

PulseDiagAgent

GPT-4o · Chain-of-Thought · MEDDxAgent (ACL 2025)
AI Differential Diagnosis Engine

Differential diagnosis engine based on the MEDDxAgent framework (NEC Research, ACL 2025). Generates ranked diagnoses with step-by-step clinical rationale using Chain-of-Thought prompting.

Patient Profile Construction — Demographics + symptoms assembled into structured prompt
GPT-4o Inference — Chain-of-Thought reasoning with temperature=0.0 for deterministic output
Format Recovery — MEDDxAgent-style retry with up to 3 attempts for robust output parsing
Ranked DDx Output — 1–20 diagnoses with rationale, confidence, and inference time
5-10sLatency
10%+Accuracy Gain
293Lines of Code
📈

PulseFlow — ED Patient Flow Forecasting

PyTorch · 2-Layer LSTM · MongoDB · MLflow
ED Trolley Count Forecasting Dashboard

LSTM neural network predicting Emergency Department patient flow across M.S. Ramaiah network hospitals. Uses 7-day feature windows with autoregressive forecasting for 1–14 day predictions.

Model Loading — LSTM checkpoint (210KB) with GradientBoosting fallback + demo mode
Feature Extraction — 7-day × 5-feature sequences (patient count, admissions, discharges, elderly waiting)
Autoregressive Forecast — Day-by-day predictions fed back as input, with confidence intervals
<50msInference
12Hospitals
24,210Records
210KBModel Size
💊

CarePlanPlus — Treatment Pathways

BERT-base-uncased · 96-Class ICD-10 · MIMIC-IV

Fine-tuned BERT model (439MB) classifying ICD-10 procedures from diagnosis sequences. Sequential pathway prediction considers prior procedures for context-aware recommendations.

Pathway Construction — Diagnosis → procedure sequences from MIMIC-IV patient records
BERT Inference — Frozen first 6 layers + custom classification head (768→256→96)
Iterative Recommendation — Sequential procedure prediction with satisfaction-weighted confidence
Similar Case Retrieval — Jaccard similarity matching from 215-patient database
100-300msLatency
96Procedure Classes
215Patients
439MBModel
📋

PulseNotes — Clinical NLP & RAG

Bio_ClinicalBERT · FAISS IndexFlatL2 · 768-dim

Semantic search engine over 1,203 clinical notes using Bio_ClinicalBERT embeddings and FAISS vector indexing. Captures medical context — not just keywords — for instant retrieval across discharge summaries, nursing notes, and radiology reports.

Index Building — Bio_ClinicalBERT [CLS] token pooling → 768-dim vectors → FAISS L2 index (69MB)
Query Processing — Natural language → embedding → top-K nearest neighbours by L2 distance
Result Assembly — Ranked results with relevance scores, patient IDs, categories, and full note text
50-200msSearch
1,203Notes
768-dimEmbeddings
🏥

MediSync — Hospital Simulation

MADDPG · MAPPO · WebSocket · 9 Departments
Multi-Agent RL Hospital Resource Optimization

Multi-agent reinforcement learning for hospital resource optimization. Simulates 9 departments with 4 staff types, using both MADDPG and MAPPO algorithms with 5-stage curriculum learning.

Environment Setup — 9 departments × 4 staff types, realistic patient arrival simulation
Curriculum Training — 5-stage progressive difficulty (Easy → Full), 168-hour episodes
Live WebSocket — Real-time streaming of department metrics during active simulation
~5msPer Step
9Departments
2RL Algorithms
💬

MSR Pulse Dashboard

React 19 · TypeScript 5.9 · Vite 8 · Recharts 3

Clinical operations dashboard with 6 interactive pages — from multi-hospital patient flow comparison charts to a session-based agent chat with approval gates for diagnosis and treatment results.

Dashboard4 stat cards, patient flow chart, recent queries
DiagnosisSymptoms form, CoT toggle, ApprovalGate
ForecastMulti-hospital comparison, breach alerts
Clinical NotesSemantic search, expandable results
Simulation3×3 dept grid, live sparklines
Agent ChatSession-based, markdown, route badges
1,950Lines TypeScript
6Pages
5Health Dots (30s poll)
Why MSR Pulse

Platform Advantages

The only platform combining all 6 clinical AI capabilities in a single deployable stack — with research-backed diagnosis, human-in-the-loop safety, and enterprise-grade infrastructure for Indian hospital operations.

🎯 Unified Interface

Single natural-language entry point for 6 AI services — eliminates the need for clinicians to learn 6 different UIs. One question, all capabilities.

🔒 Human-in-the-Loop

Mandatory clinician approval for diagnosis and treatment recommendations. The AI never presents as final authority — clinical decisions are blurred with CSS until acknowledged.

📊 Research-Backed

PulseDiagAgent implements the MEDDxAgent framework (ACL 2025) — showing 10%+ accuracy improvement with Chain-of-Thought explainability.

💰 Dual-LLM Cost Savings

GPT-4o-mini for routing (10× cheaper) + GPT-4o only for complex reasoning — reduces API costs by approximately 60%.

🧪 Real Clinical Data

MIMIC-IV dataset provides realistic patient demographics, diagnoses, procedures, and clinical notes. 24,210+ patient flow records from M.S. Ramaiah network hospitals.

🚀 Production-Ready

Docker, Kubernetes (GKE), CI/CD with GitHub Actions, Prometheus monitoring, Grafana dashboards, structured JSON logging — enterprise-grade infrastructure.

⚡ Faster Diagnosis

Explainable differential diagnosis in 5-10 seconds vs. manual review. 1-14 day ED trolley forecasting with breach risk indicators.

🔬 Multi-Algorithm RL

Both MADDPG and MAPPO for hospital optimization with 5-stage curriculum learning — enables comparison and selection per scenario.

AI Capability Performance

Differential Diagnosis5-10s response
ED Patient Flow Forecasting<50ms inference
Treatment Recommendation100-300ms
Clinical Note Search (RAG)50-200ms
Hospital Optimization (RL)~5ms per step
Query Routing~200ms (GPT-4o-mini)
Technology Stack

Enterprise-Grade Infrastructure

Built on battle-tested open-source technologies — from LangGraph orchestration to Kubernetes deployment.

LayerTechnologyPurpose
AI OrchestrationLangGraph ≥0.2.0Stateful agent workflow graph with 11 tools
LLM ProviderOpenAI GPT-4o / GPT-4o-miniReasoning, routing, diagnosis
ML FrameworkPyTorch (Latest)LSTM, BERT, MADDPG, MAPPO
NLPHuggingFace TransformersBERT, Bio_ClinicalBERT embeddings
Vector SearchFAISSSemantic similarity search over clinical notes
API FrameworkFastAPI / FlaskRESTful microservice APIs
FrontendReact 19 + TypeScript 5.9Clinical operations dashboard
DatabaseMongoDB 6.0Centralized data store
MonitoringPrometheus + GrafanaMetrics scraping + dashboards
MLOpsMLflow v2.9.2Experiment tracking, model registry
ContainersDocker / Kubernetes (GKE)Containerization & orchestration
CloudGoogle Cloud (GKE Standard Zonal)ap-south-1 (Mumbai) cluster deployment
Security & Safety

Clinical Safety & Compliance

Every clinical decision passes through mandatory approval gates. The AI never presents as final authority.

🔒 Approval Gates

Diagnosis and treatment results are blurred with CSS filter: blur(6px) and require clinician acknowledgment before viewing. The ApprovalGate component enforces this on every clinical response.

🔑 Secret Management

No secrets hardcoded in source code. OPENAI_API_KEY via environment variable only. Kubernetes secrets not checked into Git. .env.example template provided.

⚡ Iteration Safety

Maximum 15 iterations enforced as safety guard against infinite tool-call loops. Agent forced to validate regardless of pending calls.

♿ WCAG 2.1 AA

High-contrast text, focus rings on all interactive elements, semantic HTML, and aria-hidden for decorative elements. Accessible to all clinicians.

🛡️ Error Resilience

Tool failures caught and returned as messages. MongoDB unavailability triggers graceful degradation. OpenAI rate limits handled with tenacity exponential backoff.

🔐 DPDPA 2023 Compliance

India's Digital Personal Data Protection Act 2023 compliance built-in. PHI anonymization requirements. Data retention policies. MIMIC-IV requires PhysioNet credentialing + CITI training. Data sovereignty ensured via ap-south-1 (Mumbai) deployment.

Market Position

Competitor & Market Analysis

MSR Pulse is the only platform combining all 6 capabilities — diagnosis, forecasting, treatment, note search, resource optimization, and agentic orchestration — in a single deployable stack for Indian hospital operations.

PlatformFocusMSR Pulse Differentiator
Epic Cognitive ComputingEHR-integrated CDSProprietary & tightly coupled to Epic EHR; MSR Pulse is EHR-agnostic
IBM Watson HealthOncology treatmentNarrow cancer focus; MSR Pulse covers 6 clinical domains
Google MedPaLMMedical Q&AResearch-only, no deployment tooling; MSR Pulse is production-ready
John Snow LabsClinical NLPNLP-only; MSR Pulse integrates NLP + forecasting + RL + diagnosis
Nuance DAX / MicrosoftAmbient documentationDictation focus; MSR Pulse provides decision support
Viz.aiStroke detectionImaging-only; MSR Pulse works with text-based clinical data
6-in-1Only integrated platform with all 6 clinical AI capabilities
100%Open-source friendly infrastructure (Docker, K8s, MongoDB)
12Partner hospitals in M.S. Ramaiah network
ACL 2025Research-backed MEDDxAgent framework

Experience the Future of Clinical AI

MSR Pulse combines 6 specialized AI services under a single LLM-powered orchestrator — production-ready with Docker, Kubernetes, and enterprise-grade monitoring for M.S. Ramaiah hospitals.

← Back to Prana MSR Health Partner With Us View Products