• Designed and built scalable backend services for real-time data synchronization, processing 10M+ records across distributed systems using FastAPI, SQLAlchemy, Redis, and Kubernetes, enabling reliable data access for 1M+ daily active platform users.

• Diagnosed and resolved critical AWS DMS data truncation affecting 37-65% of production records, implementing a custom bi-directional data sync pipeline that restored data integrity for 50M+ record operations.

• Reduced data sync pipeline runtime by 40% through chunked batch processing and built CloudWatch monitoring infrastructure with automated alerts, enabling near-real-time data availability.

Course Producer for CSCI572 - Information Retrieval and Web Search Engines

• Designed comprehensive taxonomy framework for logical fallacy classification and developed knowledge graph injection methods for transformer models, achieving 20% performance improvement in fallacy detection.

• Contributed to dialogue modeling research for negotiation strategy identification, implementing multi-task learning approaches with LLMs that improved cross-domain performance by 25%

• Built a low-latency gaming platform serving 1M+ registered users and 100K concurrent users, covering OAuth authentication, real-time leaderboards, and full-stack development across Node.js, TypeScript, React and MongoDB.

• Developed ML-powered recommendation system for scanning regulatory compliance documents and answering customer queries using transformer embeddings and Elasticsearch, reducing customer inquiry response time by 40%.

• Designed similarity search pipeline incorporating clustering, approximate nearest neighbors, and optimized distance metrics for efficient query ranking at scale.