Case Studies & Architectural Impact

Transforming complex clinical and enterprise requirements into resilient, compliant systems.

Curio Digital Therapeutics (Current Role)

HIPAA-Compliant Real-Time AI Inference Engine

Architecture Lead

The Problem

The clinical platform required real-time AI processing capabilities to support digital therapeutics, but integrating Large Language Models natively posed extreme risks of leaking Protected Health Information (PHI) to third-party endpoints.

The Constraints

Strict HIPAA compliance. Zero tolerance for PHI leaks. The system had to support burst traffic without degrading the core application performance.

Architecture & Solution

Designed a privacy-first Retrieve-and-Generate (RAG) pipeline. Architected a critical PII/PHI Redaction Layer using Microsoft Presidio, executing before any LLM inference calls to guarantee zero data leakage. Furthermore, I decoupled the heavy inference operations from the main application thread using an asynchronous, event-driven pattern via Apache Kafka, ensuring the UI remained completely responsive during processing.

Outcome & Impact

Deployed a fully compliant AI engine that met all regulatory clinical constraints. The event-driven architecture successfully absorbed 100% of burst traffic volatility, maintaining core system stability.

Impact: Zero-PHI leakage; 400ms inference time.

LangChainPresidioApache KafkaAWSPythonTypeScript
Curio Digital Therapeutics (Consulting via SmartData Inc)

Decomposing a Legacy Monolith (Strangler Fig Pattern)

Lead for Migration / Architecture Lead

The Problem

A high-traffic legacy monolithic system was becoming fragile, difficult to scale, and painfully slow to update, bottlenecking the engineering organization's ability to release new clinical features.

The Constraints

The system could not be taken offline for a "big bang" rewrite. High data integrity requirements.

Architecture & Solution

Led a surgical migration strategy utilizing the Strangler Fig pattern. Introduced an Anti-Corruption Layer (ACL) to shield new microservices from legacy technical debt. Implemented Change Data Capture (CDC) to synchronize state seamlessly between legacy SQL stores and the newly deployed distributed microservices.

Outcome & Impact

Reduced deployment risk significantly (effectively zero-downtime migration). Restored team velocity and provided a scalable architectural foundation capable of supporting future growth.

Impact: Microservice scaling + event architecture.

Strangler FigCDCAnti-Corruption LayerAzureNode.js
Reinsurance Group of America (Consulting via TCS)

Enterprise Portal Revamp for High-Volume Data

Offshore Team Lead / Lead Architect

The Problem

The Arius Portal, a critical enterprise application, had become outdated and suffered from significant performance bottlenecks, causing friction for end users relying on rapid data access.

The Constraints

Required highly secure, federated access spanning multiple international time zones. Required strict coordination across large offshore and onshore development teams.

Architecture & Solution

Led a diverse, remote engineering team in a complete redesign of the backend architecture. Directed the modernization of authentication by integrating AWS Cognito and Okta, establishing a pristine, federated access management system. Implemented rigorous CI/CD pipelines to automate testing and deployments, establishing a modern DevOps culture.

Outcome & Impact

Achieved a measured 25% improvement in overall portal performance. Strengthened security posture to exceed enterprise audit requirements. Delivered all project milestones ahead of schedule despite complex cross-timezone dependencies.

Impact: 25% performance improvement.

Node.jsReactAWSDockerOkta/Cognito
Goods and Services Tax (GST) Portal (Consulting via Infosys)

Automated Fraud Detection Integration at National Scale

Team Lead

The Problem

The national taxation portal was highly susceptible to fake business registrations via fraudulent physical addresses, resulting in significant administrative overhead and lost revenue compliance.

The Constraints

National scale infrastructure requirements. Absolute accuracy necessary—false positives would punish legitimate businesses.

Architecture & Solution

Directed the architectural integration of MapMyIndia's spatial and location APIs directly into the core GST registration workflow. Engineered custom cross-referencing capabilities to map declared business addresses against verified geographical databases in real-time.

Outcome & Impact

Successfully automated the detection and exclusion of fraudulent registrations at the point-of-entry, drastically reducing manual verification workloads and significantly increasing system trust and legitimate tax compliance.

Impact: Distributed state architecture.

AngularNode.jsMapMyIndia APIs

Facing complex backend or AI integration challenges?

Let's Talk Architecture