Final-year B.Tech CSE (AI & ML) student at VIT-AP University.

I build backend systems, AI/ML applications, and GenAI infrastructure with a focus on scalability, reliability, and performance.

• Open-source contributor to Librosa, Gradio, LlamaIndex, LangChain, and TheAlgorithms
• Built a Deepfake Audio Detection system trained on 78,000+ audio segments
• Engineered enterprise retrieval systems indexing 10,000+ documents with sub-100ms latency
• Developed multi-agent AI architectures using LangGraph
• Solved 900+ Data Structures & Algorithms problems
• Experienced with Java, Python, FastAPI, Machine Learning, RAG, and Agentic AI systems

• Python
• Java
• SQL

• Machine Learning
• Deep Learning
• PyTorch
• LLMs
• RAG
• LangChain
• LangGraph
• LlamaIndex

• FastAPI
• Flask
• REST APIs
• PostgreSQL
• Vector Databases (FAISS, Chroma, Pinecone)

• AWS
• Docker
• Git
• GitHub Actions
• CI/CD
• Pytest
• Unit Testing

Java • Multithreading • Concurrency • System Design

High-performance financial exchange simulation implementing heap-based order books, price-time priority matching, and thread-safe order execution.

PyTorch • Deep Learning • Audio AI

Fine-tuned IndicWav2Vec on 78,000+ Hindi audio segments, achieving 0.8293 AUC on speaker-disjoint evaluation.

React • Node.js • LLMs

AI-powered code review platform capable of vulnerability detection, architectural analysis, and context-aware project understanding.

LlamaIndex • FAISS • FastAPI

Scalable retrieval platform managing 10,000+ document embeddings while maintaining sub-100ms retrieval latency.

LangGraph • Agentic AI • Multi-Agent Systems

Implemented 23 advanced agent architectures including semantic memory, guardrails, Text-to-SQL, observability, semantic caching, and agentic RAG.

FastAPI • FAISS • Backend Engineering

Semantic caching AI proxy that reduces redundant LLM requests, lowers inference costs, and improves response latency.

• Resolved a race condition in the Gallery component affecting image selection state.
• Optimized Svelte reactive logic to correctly handle dynamic image appends.
• Merged PR #13178

• Implemented hybrid_tempogram based on Peeters (2005) for advanced rhythm analysis.
• Navigated 22 commits and 59 review discussions before merge.
• Merged PR #2004

• Fixed ChatOllama tool-calling crash caused by malformed streaming JSON.
• Added robust error handling and recovery mechanisms.
• Active PR #35201

• Prevented configuration corruption from empty or whitespace-only values.
• Added comprehensive unit tests following AWS standards.
• Active PR #10003

• Investigated memory management behavior and proposed architectural improvements during maintainer reviews.

• Implemented mathematical and geometric algorithms in Java and Python.
• Added tests, CI validation, and repository-standard compliance.
• 3 PRs Merged