Shipping production AI on top of battle-tested backend infrastructure
AI Engineer with a production backend foundation — building LLM-powered systems using LangChain, LangGraph, and RAG pipelines on top of scalable Java/Spring Boot infrastructure.
Hands-on with agentic workflows, vector databases, hybrid retrieval, and LLM evaluation (RAGAS, LangSmith). Currently at Predii extending LLM retrieval systems and building AI-driven data pipelines.
My backend engineering background — microservices, event-driven systems, async processing — gives me an edge in shipping AI that actually works in production, not just in notebooks.
LangChain, LangGraph, RAG Pipelines, Qdrant, BM25, Vector Databases, LangSmith, RAGAS, Pydantic, Streamlit
Ollama, Google Gemini, HuggingFace
Python, Java, JavaScript, TypeScript, SQL
Spring Boot, FastAPI, Node.js, NestJS, Microservices Architecture
PostgreSQL, MongoDB, MySQL, Redis, Qdrant, Azure Cosmos DB
RabbitMQ, Apache Kafka, Event-Driven Architecture
Git, Docker, CI/CD, JMeter, Load Testing
JUnit, Mockito, Integration Testing
Predii, Pune
LogiQuad Solutions, Pune
VIT Pune
Team Quark, VIT Pune
Python · LangChain · FastAPI · Qdrant · BM25 · HuggingFace · LangSmith · RAGAS · Chrome Extension
RAG pipeline over YouTube transcripts using hybrid retrieval (BM25 + dense vector search), query expansion, and cross-encoder reranking. Delivers timestamp-grounded answers via a Chrome Extension embedded in YouTube UI. Evaluated with RAGAS (Faithfulness: 0.75, Relevancy: 0.71) and traced end-to-end via LangSmith.
Python · LangGraph · LangChain · FastAPI · Google Gemini · Ollama · Pydantic · LangSmith
GitHub App that reviews pull requests using a parallel multi-agent LangGraph pipeline — security, bug risk, and performance agents run simultaneously with typed Pydantic outputs aggregated into a structured merge recommendation. Supports Ollama and Gemini 2.0 Flash with a configurable human-in-the-loop gate, LangSmith tracing, and SQLite checkpointing.
View my complete resume for a detailed overview of my experience, skills, and achievements.
View ResumeSharing insights on AI engineering, LLMs, and backend systems
Deep dive into the most powerful Java 8 Stream operations. Learn how filter, map, and reduce can transform your data processing code with practical examples and real-world use cases.
Read on MediumComprehensive guide covering all major Java 8 features including Streams API, Lambda expressions, Optional class, and functional interfaces. Essential reading for every Java developer.
Read on MediumSynchron's Stentrode is a minimally invasive brain-machine interface that allows people with paralysis to control devices using their thoughts — a breakthrough in neuroprosthetics.
Read on MediumVishwakarma Institute of Technology (VIT), Pune
CGPA: 8.6/10
Relevant Coursework: Data Structures & Algorithms, Object-Oriented Programming, Database Management Systems, Computer Networks, Operating Systems, Artificial Intelligence, Cyber Security, Compiler Design
Open to full-time AI engineer opportunities and interesting collaborations