I've built machine learning systems for source-code analysis at IIIT Dharwad, using CodeBERT, β-VAE, and Random Forest for anomaly detection and authorship attribution.
I developed semantic code embeddings and anomaly-detection workflows in Python with PyTorch and Transformers, and generated Falcon-RW-1B code obfuscations to test robustness against adversarial transformations. My evaluation across Java, Python, and C++ datasets achieved 86% accuracy using ROC-AUC, Precision, Recall, and F1-score.
I co-authored and presented research at IEEE CICT 2025. I also earned second place in a university hackathon by building a beginner-friendly portfolio hedging and risk-analysis tool with CSV uploads, Random Forest predictions, dashboards, and recommendations.
Alongside my technical work, I've built an automated street-lighting prototype with C, Arduino IDE, ESP8266, PIR, and LDR sensors. I bring teamwork and creativity from competing in football and leading music activities for my college cultural club.
