Bahar Afshar

Background

About Bahar Afshar

I am a research-focused ML Engineer with a strong software development background, passionate about solving real-world problems at scale. I’m completing my Master’s in Computer Science at the University of Ottawa (graduating September 2025) and currently doing a MITACS internship at H3M Analytics under Dr. Paula Branco’s supervision. My thesis work includes developing X-ITERADE, an explainable unsupervised anomaly detection framework that detects high-quality suspicious cases in extremely imbalanced financial datasets, improving the imbalance ratio by 15x without labeled data. I also designed ALISA, a novel pipeline that leverages weighted active learning, semi-supervised learning, and data augmentation to boost fraud detection F1 scores by 22% in challenging one-class settings. Additionally, I implemented an explainability module using LLaMA, Mistral, and GPT models with prompt tuning, achieving 90% accuracy in group behavior analysis. I’ve published 3+ peer-reviewed papers on anomaly detection, active/semi-supervised learning, and explainability in finance in collaboration with H3M Analytics. Technically, I’m skilled in Python, PyTorch, Pandas, Scikit-learn, and cloud-based ML workflows, with experience in Docker, Kubernetes, and distributed systems. I’m eager to contribute as a Machine Learning Engineer and open to relocation within Ontario.