My research on computationally efficient deep federated learning for IoT botnet attack detection achieved 99.93% accuracy and F1 score. I used DP-Adam, MLP, FedAvg, XGBoost, and PCA, and the work was published in Intelligent Systems with Applications.
At Nethermind, I supported research on optimizing zero-knowledge proofs for LLaMA 2 inference and implemented verifiable XGBoost fraud detection using EZKL. As a freelance Web Scraping Specialist at Mindrift, I built an asynchronous Python pipeline to extract and validate real estate agent data from JavaScript-heavy platforms.
