Transformer-Based IoT Cyberattack Detection
Built a GPT-2 based transformer classifier for cyberattack detection on Edge-IIoT network data. Used BPE tokenization and contextual embeddings for sequential feature modeling. Achieved approximately 95% classification accuracy through architectural tuning and feature engineering. Technologies: Python, PyTorch, Transformers, Scikit-learn.
Shazam - Song Recognition Algorithm
Implemented a Shazam-like audio fingerprinting system using spectrogram peak constellation maps. Designed a DSP pipeline with WAV preprocessing, STFT-based spectrogram computation, and peak detection. Technologies: Python, NumPy, SciPy.
Mental Health Support Dashboard - OSMI Tech Survey
Developed an ML-based treatment recommendation system with confidence-aware predictions. Applied clustering-based persona analysis and interactive exploratory data analysis. Built a deployable Streamlit dashboard for real-time inference and visualization. Technologies: Python, Streamlit, Pandas, Scikit-learn.