Software Engineer II, QuickBooks Platform at Intuit

Architected a real-time financial data synchronization service in Python using Kafka event streams and Cassandra for state persistence, reducing transaction-sync propagation latency by 38% across 12 regional data centers serving 8M+ QuickBooks Online users. Designed and deployed an agentic AI bookkeeping assistant using LangChain agents with internal MCP tooling, improving automated categorization accuracy by 14% for 3 high-volume SMB cohorts in 8 weeks. Built a GenAI-powered expense forecasting microservice on AWS SageMaker with Lambda inference endpoints, cutting cold-start latency by 52% and reducing over-accrual events by 22% across TurboTax Business filings. Migrated 6 legacy Python monolith services to FastAPI-based microservices on AWS EKS, eliminating 3 critical SLA breaches per quarter and reducing infrastructure spend by $180K/year. Implemented distributed tracing using Jaeger and Prometheus across 18 downstream services, cutting MTTD by 45% and reducing on-call pages by 30%.

Software Engineer II at Razorpay

Owned backend architecture for Razorpay’s Smart Collect reconciliation service using Python (FastAPI) and PostgreSQL, leveraging distributed systems patterns such as event sourcing and idempotent retry queues; processed 4M+ daily transactions with 99.97% uptime over 18 months. Delivered an end-to-end payment routing ML pipeline: Python feature engineering on Apache Spark, trained a gradient-boosted classifier, and served predictions via Flask API; improved routing success rate by 9.2%, translating to $1.4M recovered GMV annually. Designed multi-cloud failover strategy across AWS and GCP for critical payment APIs, achieving RPO of 30s and RTO of 90s and eliminating 2 regional outage events in H2 2022. Built automated fraud-signal ingestion pipeline using AWS Kinesis and a Redis caching layer, reducing fraud-check p99 latency from 320ms to 48ms and blocking 12K+ fraudulent transactions in the first 60 days.