My Projects
Jan 2024 - Mar 2024
Implemented text classification using Bidirectional LSTM with pre-trained GloVe embeddings, achieving 91% accuracy on AG News dataset and 57% on the more complex 20 Newsgroups dataset with 20 categories.
Nov 2023 - Feb 2024
Developed an AI-powered ad generation platform that creates personalized video ads in real-time based on user preferences, integrating OpenAI for scripts, ElevenLabs for voiceovers, and D-ID for lip-synced videos.
Oct 2023 - Dec 2023
Built Part-of-Speech tagging models using both traditional ML (Logistic Regression, SVM) and deep learning (BiLSTM) approaches. The BiLSTM model with GloVe embeddings achieved 94.78% accuracy, significantly outperforming traditional methods.
Oct 2023 - Oct 2023
Created a personalized academic planning platform that analyzes 10,000+ courses across 200+ majors to deliver tailored recommendations based on student preferences, improving course selection accuracy by 40%.
Aug 2023 - Dec 2023
Built a full-stack e-commerce app with secure authentication, cart management, and responsive design. Improved page load times by 20% through efficient content rendering using Thymeleaf templates.
Jun 2023 - Aug 2023
Analyzed cross-impact dynamics of Order Flow Imbalance (OFI) in equity markets, integrating multi-level OFI metrics and applying dimensionality reduction to evaluate their predictive power for price changes across various stocks.
May 2023 - Jul 2023
Compared 3D CNNs, SVMs, and Random Forests for classifying MRI scans into PD, Control, SWEDD, and Prodromal groups. Random Forest achieved the highest accuracy (65.9%) for multiclass classification, while SVM performed best (76.4%) for binary classification.
May 2023 - Jul 2023
Developed an adaptive learning algorithm to adjust rates based on rejects, improving neural network reliability by 90% for safety-critical applications. Tested on breast cancer datasets, synthetic data, and MNIST, achieving a 10% accuracy improvement.
2020 - 2021
Designed and developed a robotic system for inspecting pipelines, capable of navigating through various pipe diameters and detecting structural issues. Awarded Indian Design Patent: 360985-001.
2019 - 2020
Designed an electric bike specifically for senior citizens with enhanced safety features and ergonomic design. Awarded Indian Design Patent (Number: 344685-001).
2020 - 2021
Developed an automated system for public ration distribution to improve efficiency and reduce fraud. Received Indian Copyright Patent (Number: SW - 14713/2021).
Oct 2020
Applied CNNs and deep learning models for accurate classification of Cats vs. Dogs, achieving 92% accuracy. Implemented using TensorFlow and Keras with data augmentation techniques.
Apr 2021 - Jun 2021
Built a voice-activated Android application for display panel printing, integrating Bluetooth communication and Arduino-based hardware controls. Improved operational efficiency by automating manual processes and reducing errors.
Jun 2022 - Aug 2022
Developed an AI-powered image processing system using YOLOv5 to detect fire, smoke, and face masks in real-time. Integrated OpenCV for preprocessing and optimized the model for low-latency deployment in safety-critical applications.