Projects
This section showcases a selection of my key projects, highlighting my expertise in deep learning, natural language processing, and data analysis. Each project reflects my ability to tackle complex challenges, optimize performance, and innovate solutions in various technical domains. From enhancing AI security to developing user-friendly web applications, these projects demonstrate my technical skills, collaborative efforts, and commitment to excellence. Explore the descriptions below to learn more about my contributions and achievements in each project.
Summarizer Web-App for Call Log Summarization
Developed an NLP-powered web app using Mistral Instruct LLM to summarize call logs and answer questions based on input dates. Integrated FastAPI and Gradio UI for seamless functionality and efficient data storage with SQLite.
Stealthy Syntactical Backdoor Attack on Language Models
Proposed a novel backdoor attack on language models, increasing attack success rate by 23%. Implemented a robust defense mechanism using GPT-3, reducing attack success rate by 50% across various models and datasets.
Comprehensive Analysis of Google Play Store Apps & Reviews
Performed extensive analysis on a dataset of 2.3 million Google Play Store entries using Python and PySpark. Uncovered insights into app popularity and user engagement, achieving 80% test accuracy in predicting 'Maximum Installs'.
Cross-Architectural Self-Supervision for Multi-Modal Learning
Developed CASS-MM, a novel technique for classifying memes in Facebook AI’s hateful meme dataset. Achieved superior performance, with a 10% accuracy improvement over CLIP-trained models.
