Experience

Key Responsibilities

• Lead a Kafka-to-S3 reconciliation pipeline handling 1B+ events/day with AWS Lambda, powering fraud models with <5-min SLA.
• Built an LLM-powered Payments Agent microservice (Python + FastAPI + RAG) projected to generate $1M+ ARR.
• Engineered user state tracking via Kafka & Java to support churn prediction, reducing drop-off by 12%.
• Developed a React-based one-click onboarding flow, lifting Instant Deposit adoption by 55%.
• Built E2E test suite with Playwright and React Testing Library to ensure frontend reliability.
• Enforced role-based access with OPA (Rego), boosting SMB customer engagement.
• Integrated Amplitude tracking for user behavior and product analytics.

Technologies Used

Key Achievements

• Handling 1B+ events/day with <5-min SLA
• Built microservice projected for $1M+ ARR
• Reduced user drop-off by 12%
• Increased Instant Deposit adoption by 55%

Key Responsibilities

• Developed a React-based deposit list and tracker with filtering & download, improving satisfaction.
• Built backend APIs using Spring Boot to manage deposit statuses, reducing operational issues.

Technologies Used

Key Achievements

• Improved customer satisfaction with deposit tracking
• Reduced operational issues through better API design

Key Responsibilities

• Built an offline eval framework (Java, JUnit) to test search ranking models over 200M queries.
• Prototyped new engagement features with clickstream signals, improving nDCG@10 by 5%.
• Refactored debugging tools using Java & Kata, improving modularity and team debug efficiency by 20%.

Technologies Used

Key Achievements

• Tested models over 200M queries
• Improved nDCG@10 by 5%
• Increased debug efficiency by 20%

Key Responsibilities

• Led MLOps pipeline development with MLflow, Docker, and Kubernetes to streamline model iteration.
• Co-developed a BERT + VGAE belief embedding model, boosting accuracy by 18%.
• Built scalable scrapers with Selenium + Redis for social media data (100K+ records).
• Developed React-based visualization dashboard for 2D/3D influence graphs.

Technologies Used

Key Achievements

• Streamlined model iteration with MLOps pipeline
• Boosted model accuracy by 18%
• Collected 100K+ social media records
• Built visualization dashboard for influence graphs