Projects
-
- Developed a multimodal RAG assistant using LangGraph, integrating Qwen2.5-VL for image
understanding, LightRAG for knowledge graph construction, and Claude 3 Sonnet for
context-aware response generation.
- Integrated the Model Context Protocol (MCP) to enable live web search, YouTube
transcript summarization, and structured data extraction, supporting comprehensive multi source document
analysis and question answering.
- For demonstration designed and deployed a scalable Frontend with Streamlit, Backend with Django REST
Framework, and FastAPI, enabling efficient workflow orchestration, session persistence, and execution
monitoring.
-
- Built a browser based C debugger supporting compilation and interactive execution.
- Integrated GDB MI through a C++ execution engine to expose program state, including
stack frames, local variables, and standard output.
- Developed a FastAPI middleware with REST and WebSocket endpoints for low-latency
communication and isolated user sessions.
- Implemented a responsive React interface with execution history and real-time
visualization of debugging state.
-
- Developed a real time particle simulation incorporating inverse distance squared repulsive force
interactions between particles using SFML.
- Benchmarked execution time and frame rates, analyzing trade-offs between CPU and GPU approaches.
- Implemented on both CPU and GPU(Cuda) to demonstrate the differences in performance between CPU and GPU
at IISc Open Day 2026.
-
- Built a C++ expression evaluator tool for Eigen matrices, supporting efficient evaluation of algebraic
and element-wise matrix operations, supporting all standard binary and unary operators(i.e., Addition,
Logarithm, Exponentiation, etc.).
-
- For handling arbitrarily large numbers beyond 264 in C++, supporting basic arithmetic
operations.
- Designed and implemented custom data structures and algorithms from the ground up to efficiently handle
computations involving ultra large integers, using the basic principles of object-oriented programming.