AI-Assisted Development

Modernized Web Portal for Healthcare (PBM) with AI-Powered UX

Healthcare PBM Organization

Overview

Designed and delivered a modern, modular web portal for a large-scale PBM organization — replacing legacy reporting interfaces with a microfrontend architecture that supports per-customer feature enablement, AI-powered search and navigation, and role-based access across complex PBM hierarchies.

Client Profile

IndustryHealthcare / PBM + Digital Member & Client Services

RegionNorth America

HeadquartersUnited States (East Coast)

OperationsNationwide (members, pharmacies, employer plans)

Company SizeEnterprise PBM / Health Services Platform

Core MissionProvide digital self-service access for PBM clients and internal operations teams to manage claims, members, plan configurations, and reporting securely and at scale.

Key Services

Client & Operations PortalRole/Hierarchy-Based AccessModular Product DeliveryM2M Integrations

The Challenge

Creating a pluggable base architecture with independent module deployment

Unified authorization for both human users and M2M integrations

Granular access control aligned to PBM hierarchy (payer > employer group > plan > member)

Need for AI-powered search and intelligent navigation across millions of records

Solution Architecture

Implemented microfrontends with Single SPA — each module (Claims Search, Formulary Lookup, Reporting) developed as a standalone React app with its own CI/CD pipeline.

Built reusable API Gateway custom authorizer validating JWTs, loading permissions from MongoDB. Designed dual authentication model for human users (Cognito) and machine clients (OAuth2).

Integrated AI-powered search and smart filtering for rapid claim and member lookups across large datasets.

Architecture Diagram — Modernized Web Portal for Healthcare (PBM)

Features & Capabilities

Microfrontend Architecture

Independent React modules deployed and composed at runtime via Single SPA

Plug-and-Play Module System

Each feature is a standalone module with independent CI/CD

AI-Powered Search

Intelligent search across millions of claims and member records with smart filtering

Granular Access Control

Frontend feature-level UI visibility; Backend route-level authorization

Dual Authentication Model

Human users via Amazon Cognito (JWT); Machine clients via Cognito app clients (OAuth2)

M2M Integration Layer

Secure APIs exposed via API Gateway with consistent authentication

Tenant-Based Configuration

Module enablement, UI branding, and access rules dynamically configured per customer

Event-Driven Scalability

SQS and Kafka for async workflows

Real-Time Observability

Datadog RUM for frontend, CloudWatch for backend

Zero-Downtime Deployments

Serverless + immutable deployments

Technology Stack

Frontend Framework

React + Material UI

Architecture Pattern

Single SPA (microfrontends)

Backend Runtime

AWS Lambda (Node.js)

API Layer

Amazon API Gateway (REST)

Authentication

Amazon Cognito (users + app clients)

Hosting

S3 + CloudFront (CDN)

Messaging / Events

Amazon SQS, Confluent Kafka

CI/CD

GitHub Actions + AWS CodeBuild

Observability

Datadog RUM (frontend), CloudWatch (backend)

Permissions Store

MongoDB (hierarchical PBM roles, tenant rules)

AI/ML Layer

Intelligent search, smart filtering, recommendation engine

Security & Compliance

Identity Isolation

Users via Cognito User Pool; Machines via Cognito App Client with OAuth2

Least Privilege Access

Custom authorizer generates dynamic IAM policies based on user's PBM hierarchy

Secure Communication

TLS 1.3 enforced across all endpoints; all data encrypted at rest

No Cross-Module Leaks

Microfrontends run in isolated contexts; authorization checked independently

Compliance Alignment

HIPAA (with BAA), SOC 2 Type II, NIST SP 800-53, OWASP Top 10

Results & Impact

Feature Delivery Speed

80% faster

Customer Enablement

100% customer-specific configurations

Compliance

HIPAA-ready

Concurrent Users

Up to 10k (peak load)

Deployment Frequency

20+ deploys/month (per module)

Team Size

6-8 developers (full-stack)

Duration

1.5 years (Aug 2025 completion)

Duration1.5 years (Aug 2025 completion)

CategoryAI-Assisted Development

Team Size6-8 developers (full-stack)

Have a Similar Challenge?

We'd love to hear about your project and explore how we can help.

Start a Conversation