✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦

✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦

I'm a Metallurgical & Materials Engineering student at IIT Kharagpur who found her passion in building intelligent systems. From production-grade RAG pipelines to real-time AI agents — I love turning complex ideas into things that actually work.

I'm deeply curious about the theoretical foundations of AI and ML — from loss functions and probability theory to the math behind transformers. Hackathons are my favourite way to learn fast and build fearlessly. Currently diving into open source. apart from all this , I love to dance ✨

• Built a multi-agent system using CrewAI orchestrating specialized agents (Scout, Scribe) across three application pathways — early-stage startups, research ----professors, and big-tech roles — served through a FastAPI backend with a custom web UI. -Engineered a Playwright browser-automation tool to auto-fill simple application forms with the resume and cover letter (left for user review/submission), and integrated the Gmail API (OAuth2) to generate review-ready outreach email drafts. -Powered generation with Llama 3.3 70B via Groq and live web search via Serper; implemented sequential task pipelines with context-passing between agents and per-lane tool routing (form-fill, email-draft, or link compilation).

Production-grade RAG system for dense research documents

• Architecting hybrid retrieval combining semantic vector search (ChromaDB) with BM25 keyword retrieval via QueryFusionRetriever
• Implementing Semantic Chunking, spaCy NER, and coreference resolution to preserve document hierarchy
• Evaluating hallucination prevention using RAGAS (Faithfulness, Answer Relevance, Context Precision & Recall)
• Integrating LangSmith / Arize Phoenix for end-to-end query observability; exploring Docker containerization

Python FastAPI LlamaIndex ChromaDB BM25 HuggingFace RAGAS LangSmith

Commodity market trend forecasting · EXCAVATE Hackathon (COMPOSIT, IIT Kharagpur)

• Built a predictive ML system using material science signals to forecast commodity market volatility cycles
• Implemented XGBoost, Random Forest, and TensorFlow models with hyperparameter tuning
• Designed the full end-to-end pipeline independently — from data preprocessing to model evaluation

TensorFlow XGBoost Random Forest Python scikit-learn

Real-time conversational AI · DataQuest Hackathon (Megalith, IITKGP)

• Developed a conversational AI system for continuously evolving live data streams
• Built a RAG pipeline with Pathway for continuous ingestion, incremental updates, and dynamic indexing

Pathway RAG Python Real-time AI

GenAI-powered nutrition assistant · Hushhh AI Hackathon

• built a GenAI-powered nutrition assistant
• Engineered a FastAPI backend generating fully personalized meal plans using structured prompt engineering under multi-parameter constraints

FastAPI Gen AI Prompt Engineering Python

Languages Python (Async) SQL C C++ JavaScript

AI / ML LlamaIndex LangChain RAG Pipelines TensorFlow XGBoost ChromaDB RAGAS Vector Databases Multi-Agent Workflows Google Vertex AI

Theory & Foundations Linear Algebra Probability & Statistics Bayesian Inference Optimization Neural Network Theory Hypothesis Testing Regression Analysis

Backend & Tools FastAPI Node.js React Next.js MySQL MongoDB Docker Git LangGraph CrewAI NetworkX

• 🔨 Building LexAI
• 📖 Deep-diving into theoretical AI/ML foundations
• 🌱 Starting my open source journey with
• 💜 Learning to contribute to AI/ML tooling repos

✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦

✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦ ✦