Real Projects.
Measurable Results.

Manual medical coding and claims submission across 12 facilities caused 34% denial rates, 90-day AR cycles, and $2M+ annual revenue leakage.

Built an AI-assisted medical coding engine (ICD-10/CPT) integrated with Epic EHR via HL7 FHIR APIs. Automated eligibility verification, prior-auth workflows, and real-time denial prediction using an XGBoost model trained on 3 years of claims data.

Legacy monolithic Magento store couldn't handle 10× traffic spikes during sales events, causing 6-hour outages and $400K lost revenue per incident.

Migrated to a headless Next.js 14 storefront backed by a microservices API layer on AWS ECS. Implemented Redis-based session caching, CDN edge delivery, and an AI recommendation engine using collaborative filtering on purchase history.

Manual KYC onboarding took 5–7 business days, causing 40% customer drop-off. Regulatory AML reporting required 3 FTEs working full-time on manual data extraction.

Developed an automated KYC pipeline using OCR document extraction, liveness detection, and sanctions list screening via real-time API integrations. Built an AML transaction monitoring engine with rule-based + ML anomaly detection, generating automated SAR reports.

Zero real-time visibility into 800+ vehicles and 12,000 daily shipments. Customer complaints about ETAs were at 38% of all support tickets. Manual dispatch caused 22% empty-mile inefficiency.

Built an IoT telemetry ingestion pipeline on AWS IoT Core processing 50K events/min from GPS trackers. Developed a React Native driver app, a web dispatch dashboard, and an ML-based ETA prediction model using historical traffic and weather data.

Single-tenant monolith on bare-metal servers couldn't onboard enterprise clients requiring data isolation. Infrastructure costs were $80K/month with 60% idle capacity.

Re-architected to a multi-tenant microservices platform on Kubernetes with schema-per-tenant PostgreSQL isolation. Implemented tenant-aware API gateway, automated provisioning via Terraform, and a self-serve onboarding portal reducing sales-to-live from 3 weeks to 2 hours.

Off-the-shelf LMS couldn't support live video, adaptive assessments, and gamification simultaneously for 200K concurrent users. Video buffering complaints were at 55% of support volume.

Built a custom LMS on Next.js with WebRTC-based live classrooms, an adaptive quiz engine that adjusts difficulty using Item Response Theory, and a gamification layer (XP, badges, leaderboards). Deployed on AWS with CloudFront CDN and auto-scaling ECS clusters.