Machine Learning Intern

Diabetes Prediction System
Sep 2023– Feb 2024

• Applied advanced preprocessing techniques and explored over 10 machine learning algorithms to create a Linear Stacking Classifier, achieving superior diabetes prediction accuracy.
• Led the research from data preprocessing to model evaluation, and presented findings at an international conference, earning the Best Paper award for its innovative contributions to the field

Deep Learning Intern

Forgery Detection System via Inpainting Techniques
Feb 2025– Apr 2025

• Independently developed a computer vision system for forgery detection, achieving 95.52% accuracy, by leveraging a large publicly available image dataset, using image inpainting techniques on it and displaying the confidence score.
• Developed and deployed a DCGAN with a Streamlit UI to detect image forgeries for real-time digital forensics

Software Engineer

Commonwealth Secretariat
Jan 2025– May 2025

• Contributed to the development of an AI prototype of an Early Warning Dashboard for the Governance and Peace Directorate at the Commonwealth Secretariat.

AI-Powered CSV Analyzer Link

Built intelligent data analysis tool with natural language querying using LangChain and local LLM (Llama2). Processes datasets up to 200MB with deployed full-featured web app enabling interactive data visualization. Implemented privacy-first architecture delivering real-time insights through AI-driven pandas agents

Research Paper Summarizer and Chatbot

Built a Streamlit-based tool for uploading and querying research papers using Gemini Pro and LangChain, enabling efficient literature reviews with structured metadata extraction, CSV summarization, and smart Q&A from PDFs.

Retrieval-Augmented Generation System for Resume Analysis

Designed an resume intelligent document analysis system using RAG architecture, integrating LangChain, Ollama LLM, and vector embeddings for automated resume parsing and semantic search.

ECG-Based Arrhythmia Classification System

Conducted comparative analysis of machine learning and deep learning models for cardiac arrhythmia detection using the MIT-BIH dataset, achieving 97.32% accuracy with the Random Forest classifier.

Phishing URLs Classification System

Developed a machine learning-based Phishing URL Classification System, implementing ensemble learning techniques to accurately classify over 230,000 websites with over 50 features.