Aryan Sharma

• Developed and launched the Prediction Contest feature end-to-end, contributing to a 10% increase in revenue within the first quarter of release.
• Reduced API latency by 50% by profiling Fantasy Contest flows and optimizing queries to remove bottlenecks.
• Refactored core financial workflows, simplifying service architecture to cut infrastructure and processing costs by 50% while improving reliability.
• Introduced Redis and local in-memory caching for prediction contest flows, cutting average response times by 40%.
• Partnered with product and QA to define APIs, write technical specs, and ship zero-downtime deployments in a fast release cycle.
• Used AI-assisted development tools (Cursor, Claude Code) to accelerate feature development and bug fixes.

• Built scalable backend microservices in Spring Boot with REST APIs tuned for low-latency integrations within the TDS 2.0 tax processing system.
• Implemented Kafka-based event-driven communication between services, improving fault tolerance and enabling reliable asynchronous processing at scale.
• Developed routing and transformation pipelines with Apache Camel across 5+ heterogeneous systems, standardizing inter-service communication.
• Optimized database access with JPA/Hibernate — lazy loading, batch fetching, native query tuning — reducing average query time by ~30%.
• Achieved 85%+ test coverage with JUnit and Mockito, reducing regression bugs in production.

A Retrieval-Augmented Generation (RAG) system in Spring Boot that answers natural-language questions about a git repository’s commit history, grounded with real commit citations.

Extracts commit history via JGit into PostgreSQL, filters low-signal commits, generates OpenAI embeddings with pgvector similarity search, and synthesizes cited answers using Gemini 2.5 Flash.

An HTTP server built from raw TCP primitives — handling GET/POST requests, serving static files, and managing multiple concurrent connections via multithreading.

CGPA: 8.43 · Elective: Data Science

91%