Programming

Python, R, SQL, JavaScript

AI

Scikit-learn, LLMS, TensorFlow, PyTorch, Machine Learning Algorithms, NLP, Deep Learning Neural Networks

Data Handling

Pandas, NumPy, PowerBI, Tableau, SQL

Software Tools

Git, Jupyter, VS Code, Docker, Azure ML

SIGN LANGUAGE RECOGNITION USING CONTINUAL LEARNING

• Built a CNN-based sign language classifier using the Sign Language MNIST dataset.
• Implemented Elastic Weight Consolidation (EWC) to enable sequential learning without forgetting.
• Improved accuracy and retention while training on letter groups (A-G, H-M, N-Y).

VIDEO CLASSIFICATION WITH CNNS AND HISTOGRAM MODELS

• Developed classifiers (S3D, ResNet, Conv3D) to classify video data with temporal-spatial features.
• Improved S3D model accuracy to 60% on real-time datasets through fine-tuning.
• Evaluated models using confusion matrix and performance metrics (precision, recall, F1-score).

AI BOT WITH NANO GPT

• Built an AI bot using NanoGPT to help AI students understand concepts clearly.
• Fine-tuned on curated course materials, research papers, and FAQs for domain-specific answers.
• Generated context-aware responses with transformer-based generation.

HAND GESTURE-BASED VIRTUAL DRAWING AND TYPING SYSTEM

• Created a system to draw/type using hand gestures with fingertip tracking (MediaPipe).
• Implemented dynamic color selection and gesture-controlled keyboard.
• Enhanced UI with CV Zone and OpenCV for real-time responsiveness.

LLM-POWERED SMART Q-A SYSTEM

• Fine-tuned DistilBERT on SQuAD for contextual question answering.
• Integrated GPT-based models for dynamic, intelligent response generation.
• Used BLEU and ROUGE for LLM performance evaluation and RAG pipeline development.