Kowshik Das Ushna

I’m a final-year Computer Science and Engineering student with a strong interest in Artificial Intelligence and Machine Learning, especially in the research sector. I’m currently studying at Chittagong University of Engineering and Technology (CUET) and I’m drawn to work that turns model ideas into usable systems.

I build and evaluate AI systems end-to-end—from data processing to performance validation. My RAG-based Multi-Document PDF Chatbot uses LangChain, HuggingFace models, FAISS, and a Groq chat model, with session-based chat history for context-aware answers. I’ve also developed a CNN-based Student Attentiveness Detection model that achieved 99.87% test accuracy using PCA to reduce training time by 20%, and a CNN-LSTM Tweet Sentiment Analysis model on the Sentiment140 dataset (1.6M tweets). For computer vision applications, I designed a web-based Bokeh Effect Image Processor using TensorFlow Lite’s DeepLabV3 with OpenCV-optimized performance.

Currently, I’m working on my thesis: “Multi-Partition RAG with Dual-Agent RL for improving LLM Performance,” exploring adaptive partition selection and memory refinement. I’ve published and presented research including “CNN-Based Student Attentiveness Detection: A Hybrid Approach with Dimensional Reduction” (Atlantis Press/Springer Nature) and a QPAIN 2026 presentation on earthquake magnitude prediction. Through my industrial attachment at The Tork, I collaborated with senior engineers on Agile workflows and built an AI-powered JSON Generator via a chatbot framework for workflow automation, and I’ve also led teams as Vice President and Joint General Secretary across university and IEEE CS activities.