Engineered a clearing platform handling $15B+ daily volume with sub-300ms latency, supporting 500% client growth at 99.97% uptime. Built ML-based fraud detection reducing fraud by 28% and recovering $180,000+, while cutting false positives by 35%. Deployed LangChain document processing with 92% accuracy to reduce compliance review time by 65% and onboarding from 3 weeks to 8 days. Architected a real-time data synchronization system reducing data staleness from 15 minutes to under 5 seconds for institutional trading decisions.

Full-Stack Engineer at Osmind

March 1, 2021 - April 1, 2023

Designed HIPAA-compliant clinical dashboard reducing provider documentation time by 55% across 150+ clinics through intelligent pre-population and voice transcription. Developed FHIR API integration layer connecting 8 legacy EHR systems, reducing synchronization delays from 48 hours to under 2 hours for 40,000+ patients. Migrated Rails monolith to FastAPI + React microservices, supporting 350% user growth while reducing infrastructure costs by 30%. Implemented PyTorch treatment outcome model (74% accuracy) improving therapy success rates by 18% through earlier intervention. Automated 25,000+ monthly patient outreach via n8n workflows, improving show rates by 26% and saving 40 hours weekly.

Full-Stack Developer at Fivetran

August 1, 2019 - February 1, 2021

Built 15 data source connectors processing 2M+ daily syncs, eliminating 200+ monthly manual ETL hours. Developed schema drift detection preventing 90% of pipeline failures that previously caused 4-8 hour outages. Refactored synchronization engine to asynchronous architecture increasing throughput by 65% (12,000 to 20,000 records/second), reducing costs by $8,000 monthly. Built ML error classifier analyzing 500,000+ failures with 81% accuracy, reducing support tickets by 38%.

Full-Stack Developer at Fivetran

July 1, 2018 - July 1, 2019

Built 15 data source connectors processing 2M+ daily syncs, eliminating 200+ monthly manual hours. Developed schema drift detection preventing 90% of pipeline failures that previously caused 4-8 hour outages. Refactored synchronization engine to asynchronous architecture increasing throughput by 65% (12,000 to 20,000 records/second), reducing costs by $8,000 monthly. Built ML error classifier analyzing 500,000+ failures with 81% accuracy, reducing support tickets by 38%.