Available for full-time roles

Building context-aware
LLM applications

I build full-stack GenAI products, RAG pipelines, and reliable AI systems that turn messy model output into production-ready solutions.

View Projects Download Resume

About

Focused on software, AI systems, and the details that make them reliable

I'm a final-year B.Tech student at IIT (ISM) Dhanbad, graduating in May 2026, building full-stack and GenAI systems with an emphasis on clarity, structure, and production readiness.

I build full-stack and GenAI systems, from RAG pipelines and LLM-powered backends to interactive React frontends, and enjoy turning messy, unstructured model output into reliable, structured, production-ready systems. Currently working as a Coding & Math Annotator at Outlier AI, where I evaluate and correct LLM-generated code and math solutions, design adversarial test cases to stress-test model performance, and help improve dataset quality for LLM training and fine-tuning.

Experience

Outlier AI

Coding & Math Annotator (Remote) · January 2025 – Present

Evaluating and correcting LLM-generated algorithmic and mathematical solutions, identifying failure patterns through code analysis and debugging, and designing adversarial test cases to stress-test model performance.

Project 01

News Story Arc Tracker

A narrative intelligence platform that extracts timelines, entities, and sentiment from real-world events using AWS Bedrock LLMs. Features a scalable RAG pipeline built with LangChain, Titan Embeddings, and Pinecone/OpenSearch for semantic retrieval, an interactive React frontend with dynamic relationship graphs and data visualizations, real-time voice chat over WebSockets with streaming transcription and text-to-speech, and a FastAPI backend with streaming endpoints, model fallbacks, and reliability-focused safeguards.

React
FastAPI
AWS Bedrock
LangChain
Pinecone
WebSockets

GitHub Repo Watch Demo

news-story-arc-tracker

Project 02

Chronos: AI Story Engine

An AI-powered interactive story engine with a dual-phase streaming pipeline using Gemini 3 Flash that emits narrative text and structured JSON state in a single request. Includes a deterministic state machine that transforms unstructured LLM output into tracked game variables and story state, a lightweight RAG-style Story Card system for injecting relevant lore and memory with minimal context overhead, and a Director Mode for real-time human-in-the-loop control over tension parameters and plot flow.

React
TypeScript
Google Gemini 3 Flash
Vite
Tailwind CSS

GitHub Repo Watch Demo

chronos:-ai-story-engine

Project 03

C-Shell

A modular, Unix-inspired command-line shell built from scratch in C++ using POSIX system calls. Implements core shell functionality including command parsing, process creation via fork/exec, I/O redirection, piping between commands, signal handling, and a built-in command system — all without relying on any shell libraries.

C++
POSIX System Calls
Process Management (fork/exec)
I/O Redirection & Piping
Signal Handling
Linux

GitHub Repo

c-shell

➜~/projects/c-shellgit:(main)make

g++ -std=c++17 -O3 -Wall -c main.cpp parser.cpp shell.cpp

g++ -std=c++17 -O3 -Wall -o c-shell main.o parser.o shell.o

➜~/projects/c-shellgit:(main)./c-shell

Welcome to C-Shell v1.0.0

Type 'help' to see built-in commands.

c-shell $ls -lh

-rwxr-xr-x

1.2M

c-shell

executable

-rw-r--r--

4.8K

main.cpp

source

-rw-r--r--

8.2K

shell.cpp

source

c-shell $|

* Artistic representation. Not an actual screenshot.

news-story-arc-tracker

Contact

Let’s make it easy to reach me

Use the form for a direct message or reach out through email and social links.