Nakul saini

I’m a Computer Science undergrad with hands-on experience in Python, SQL, Machine Learning, and Deep Learning frameworks. I’m passionate about building intelligent systems and deriving meaningful insights from data, using modern AI technologies to turn unstructured information into actionable outputs.

I developed Contractezzy, a production-ready AI contract intelligence platform that automates extraction of legal terms, clauses, and obligations from unstructured documents. Using PaddleOCR, Hybrid RAG (FAISS + BM25), and GPT-4o, I built an agentic LangGraph pipeline with Reciprocal Rank Fusion to deliver structured, searchable contract insights and reduce manual review time.

I also built a Job Sarthi AI job application assistant for federal USAJobs positions, automating the recruitment workflow with a 90% output accuracy rate. Through a sequential CrewAI pipeline of specialized agents (JDAnalyst, Resume/Cover Letter Writer, and MessagingAgent), I reduced manual application time by 75% and implemented robust automated tracking.

To round out my applied ML experience, I created a used car price prediction system using a regression-based ML pipeline with scikit-learn preprocessing and a LightGBM model. Deployed with FastAPI and JWT-based authentication, it uses Redis caching and Docker Compose, while improving observability with Prometheus and Grafana.