Rakesh Purbiya

I'm building production-oriented AI and machine-learning systems as a second-year B.Tech. student at SVNIT. My work focuses on generative AI, agentic workflows, LLM applications, retrieval-augmented generation, and reliable test-driven automation.

I built an AI Software Engineering Agent using Python, LangGraph, Ollama, Docker, Streamlit, GitHub, and LangSmith. It investigates software issues, plans and reproduces bugs, applies changes in isolated sandboxes, validates tests, and prepares draft pull requests with human approval gates and checkpoint-based resume support.

I also built an Amazon Electronics recommender system with collaborative, content-based, and hybrid approaches; pure collaborative filtering achieved a HitRate@10 of 0.4046. My fraud-detection project combines an autoencoder and XGBoost across 1.8M+ simulated transactions, while my telecom churn application reached 81% accuracy and was deployed with Streamlit.

I've completed Andrew Ng's Machine Learning Specialization and continue building live ML applications across PyTorch, TensorFlow, Scikit-learn, XGBoost, and modern LLM tooling.