Arnav Raj arnav-on-code

🎓 B.Sc. Data Science & Applications @ IIT Madras — concurrently pursuing B.Tech CS @ Haridwar University
🧠 Building end-to-end ML/DL systems, LLM-powered RAG pipelines, and production-style data apps
🏢 Data Science Intern @ Central Coalfields Limited (CCL) — built a fraud detection system, +25% detection accuracy
🛠️ Comfortable across the stack: Python · TensorFlow · PyTorch · LangChain · FAISS · Hugging Face
🏆 170+ LeetCode problems solved, 50-day streak badge · State-Level Hackathon Runner-up
🎯 Currently targeting Data Scientist / ML Engineer roles — India, US, Japan & top MNCs
📫 Reach me at 23f3002537@ds.study.iitm.ac.in

🧰 Tech Stack

Languages

Machine Learning & Deep Learning

AI / NLP / LLMs

Frameworks

Cloud & DevOps

Databases

Project	Stack	Highlights
Automated RAG & Document Intelligence Pipeline	Python · LangChain · FAISS · Chroma · Ollama	End-to-end PDF/Markdown ingestion → chunking → embeddings → semantic retrieval, with BM25 hybrid search & source-citation extraction; local LLM inference for privacy-aware RAG.
AI-Powered MCQ Solver	PyTorch · Hugging Face · Sentence-Transformers · FAISS	TF-IDF/Word2Vec → BERT/BART → semantic retrieval + LoRA/PEFT fine-tuning; custom TextCNN. Kaggle RMSLE 0.759.
Heavy Industry Machine Price Prediction	Pandas · Scikit-learn · XGBoost · LightGBM · CatBoost	Full regression pipeline with feature engineering & log-target modeling; weighted ensembling. Kaggle RMSLE 0.196.
LeetCode Sync	Python · GraphQL · GitHub Actions	Automated LeetCode → GitHub sync: fetches accepted solutions, tracks stats, auto-generates README, scheduled via CI.

Data Science Intern — Central Coalfields Limited (CCL), Ranchi · Aug 2025

Built a Fraud Transaction Detection System, improving detection accuracy by 25% via precision-recall optimization
Engineered domain-specific features from transactional data; refined precision-recall balance by a further 15%

Open to Data Scientist / ML Engineer roles — let's connect!