Naman Sharma
AI Engineer • Data Scientist • IIT Guwahati
Specializing in Enterprise RAG pipelines and autonomous agentic workflows. I bridge the gap between complex data and actionable AI, turning LLM potential into production-ready intelligence.
Specializing in Enterprise RAG pipelines and autonomous agentic workflows. I bridge the gap between complex data and actionable AI, turning LLM potential into production-ready intelligence.
I am a dedicated Data Science and AI undergraduate at the Indian Institute of Technology, Guwahati, with a specialized focus on LLM Engineering, Retrieval-Augmented Generation (RAG), and autonomous agentic workflows. My passion lies in bridging the gap between theoretical machine learning and production-ready intelligent systems that solve complex enterprise challenges. I combine a robust foundation in statistical analysis and predictive modeling with advanced technical expertise in building stateful, multi-agent AI assistants..
B.Sc. (Honours) in Data Science and Artificial Intelligence | 2023 – 2027
Current CPI: 7.79
Relevant coursework includes Large Language Model Application Engineering, Deep Learning, Machine Learning Fundamentals, Time Series Analysis, Statistical Inferencing, and Relational Database Management Systems.
I engineered an enterprise-grade RAG assistant that delivers precise, grounded answers from internal documents. By building a semantic search pipeline with OpenAI’s text-embedding-3-large and ChromaDB, and integrating query rewriting, semantic retrieval, and LLM-based reranking, I minimized hallucinations while maximizing relevance. The system, powered by GPT-OSS-120B and a Gradio interface, achieved high-performance benchmarks with 96% keyword coverage, an MRR of 0.91, and an nDCG of 0.90 across 150 test queries."
I developed an autonomous AI assistant using a worker–evaluator agent architecture to execute complex, multi-step user tasks. By building stateful workflows with LangGraph, I enabled iterative reasoning and integrated powerful tools including Playwright for browser automation, web search, and Python execution. The system features a dedicated evaluator agent that verifies task completion against success criteria to ensure reliability and minimize premature responses. Deployed with a Gradio interface, the assistant facilitates interactive task submission, real-time monitoring, and seamless orchestration of data extraction and notification workflows.
I developed a machine learning model to enhance fraud detection accuracy by analyzing historical insurance claim data. The project involved comprehensive data cleaning, feature engineering, and the implementation of robust algorithms, including Logistic Regression and Random Forest. The final model achieved high prediction accuracy, enabling the early identification of suspicious claims and effectively reducing false positives.
I developed a predictive classification model to streamline loan approval decisions by identifying key borrower attributes. By performing in-depth exploratory data analysis and rigorous data preprocessing—including handling missing values and categorical encoding—I prepared the dataset for accurate model training. This project delivered a data-driven framework that facilitates faster, fairer, and more consistent decision-making processes for loan applications.
I conducted an exploratory data analysis on health insurance datasets to identify the primary factors influencing individual medical charges. By visualizing correlations between key variables such as age, BMI, and smoking status, I uncovered significant insights into cost drivers. The project concluded with the development of a regression model capable of accurately predicting individual medical insurance costs based on these demographic and lifestyle features.
I performed extensive data preprocessing and exploratory analysis to evaluate the physicochemical properties of red wine samples. By leveraging heatmaps and advanced visualization techniques, I identified key chemical features that significantly impact wine quality. The project involved training and evaluating classification models to accurately predict quality levels based on these underlying chemical attributes.
LLM Engineering, Retrieval-Augmented Generation (RAG), Agentic AI, Prompt Engineering, Semantic & Vector Search, Predictive Modeling, Machine Learning Fundamentals.
Python, SQL, Java, C, R (Elementary), LangGraph, LangChain, LiteLLM, Scikit-learn, Pandas, NumPy, Matplotlib, Gradio.
ChromaDB, OpenAI API, Git/GitHub, Power BI, Tableau (Elementary), Excel.
Online | January 2023 – July 2023
Email: dharmanaman@gmail.com
GitHub: github.com/aiwithns-ai
LinkedIn: linkedin.com/in/Naman Sharma