Harsh Kanani

Hi, I’m Harsh, a 3rd-year Computer Science and Data Science student with a passion for building AI-powered solutions that solve real-world problems. Throughout my academic journey, I’ve worked on several projects that reflect my interest in machine learning, AI systems, and scalable software development.

One of my recent projects is TrackTendance, an intelligent attendance system designed using PyTorch, OpenCV, MTCNN, and Deep SORT. It achieves real-time face recognition with 95% accuracy at 30 FPS. I also implemented an adaptive LSTM model with a clarity-based gating mechanism, significantly reducing false positives. This project gave me hands-on experience in computer vision and multi-object tracking.

Next, I developed DigInsight, an AI-powered e-commerce analytics platform using LLaMA, LangChain, and PyTorch. By building a Retrieval-Augmented Generation (RAG) pipeline, I enabled accurate trend forecasting and customer behavior analysis with 97% accuracy. I also optimized model training using techniques like gradient checkpointing and 4-bit quantization, reducing computational overhead.

Another impactful project was a Credit Card Fraud Detection System using GANs, TensorFlow, and Scikit-learn. I used synthetic data generation to improve model performance on imbalanced datasets, achieving a fraud detection accuracy of 92%. This experience deepened my understanding of adversarial networks and anomaly detection.

Additionally, I built the Reddit Data Analysis Platform using Streamlit, NLP, and LLMs. It performs sentiment analysis, credibility scoring, and topic modeling using algorithms like LDA and NMF. I integrated the Google Gemini API to generate real-time insights, helping users identify misinformation with 87% accuracy.

These projects have equipped me with a strong foundation in AI, machine learning, and software engineering. I’m always eager to tackle new challenges, explore emerging technologies, and contribute to impactful solutions.