

Data-driven professional with experience in machine learning, data analytics, and cloud computing. Proficient in Python, TensorFlow, and Tableau, with a proven ability to develop scalable AI/ML solutions for real-world challenges. Adept at collaborating in team environments to deliver innovative projects in predictive modeling, data visualization, and web application development.
DeceptionSense: Developed a real-time deception detection system using TensorFlow, Python, and OpenCV for video analysis, enhancing decision-making reliability.
ReflectEase Al: Designed a journal analysis tool with IBM Watson and NLTK to extract emotional insights, delivering a scalable mental health solution through a multi-page Flask application.
EventMingle: Engineered a Flask-based event management app with AWS for real-time coordination, implementing secure authentication and dashboards to improve user engagement.
Machine learning - TensorFlow, scikit-learn, LLM
Data analytics - Python, Tableau, NLTK
Data visualization - OpenCV, Matplotlib
Cloud computing - AWS (DynamoDB, SNS), IBM Watson
Web development - OpenCV, Matplotlib
Computer Vision
Emotion-Based Media Recommender: Developed a real-time emotion recognition system using OpenCV and DeepFace to detect user states, utilizing voice commands to deliver stabilized, personalized media recommendations.
Monetary Policy & Market Analysis : Conducted time-series analysis on NIFTY 500 data using LSTM forecasting to evaluate repo rate impacts, implementing dynamic thresholding to classify market anomalies.
TelecomQA Bot & Stock Scraper : Engineered an NLP-driven telecom support bot and stock scraper using ChromaDB, FLAN-T5, and Selenium to automate query resolution and extract real-time BSE Sensex data.
E-commerce Insights Dashboard: Designed an interactive Tableau dashboard to analyze customer behavior, implementing dynamic filters and calculated fields to identify high-value demographics and revenue trends.
Physics-Guided Solar Forecasting: Developed a Physics-Guided LightGBM framework using NASA POWER API to model aerosol-humidity interactions, achieving 94.5% accuracy to minimize grid penalties.
Digital Forensics & Vision Pipeline: Engineered a camera source attribution system using OpenCV, PRNU, and YOLOv11 to identify devices via sensor noise fingerprints, automating multi-task object detection and segmentation.