BACK TO WORK
MithilaStack2024 - PresentLead Full-Stack & AI Platform Engineer

// AI Clinic

Specialized "Agent-as-a-Service" platform that empowers healthcare providers to build, customize, and deploy AI-driven medical agents — ranging from pre-surgery counseling to chronic care management — facilitating 24/7 patient engagement and achieving 98% medication adherence.

The Challenge

Healthcare providers needed a scalable infrastructure to deploy intent-specific AI agents for patient engagement without requiring deep ML expertise — and the system had to maintain clinical accuracy, multi-language support, and compassionate interaction quality.

The Solution

Architected the core platform using Node.js, Fastify, and MongoDB, integrating LiveKit-powered real-time voice AI and a neural embedding pipeline to process medical knowledge bases into searchable clinical context. Built a multi-tenant backend supporting multiple healthcare organizations and implemented a continuity engine for automated check-ins, lab insight analysis, and appointment management.

// Key Impact Metrics

0%
Medication Adherence
0%
Admin Tasks Automated
0+
Medical Agent Types

// Tech Stack

Node.jsFastifyMongoDBLiveKitPythonNeural EmbeddingsVector StoreMulti-tenant

Key Learnings

01.

Healthcare AI agents require strict persona boundaries to prevent clinical misinformation — system prompts must encode both knowledge scope and escalation protocols.

02.

Embedding medical documents at a finer granularity (section-level vs document-level) dramatically improves retrieval precision for clinical queries.

03.

A "build once, deploy anywhere" agent factory pattern significantly reduces time-to-deployment for new healthcare provider onboarding.