Dawood Tahir

Built an end-to-end fraud detection pipeline on Databricks using SQL for large-scale feature extraction and window-based behavioral signals; conducted exploratory profiling and delivered a production-ready workflow. Developed a data-driven research workflow to extract brand-text and imagery details from websites, and designed a RAG-based inference flow to support action-oriented insights.

Built an end-to-end fraud detection pipeline on Databricks using SQL for large-scale feature extraction and window-based behavioral signals; performed exploratory profiling on 20M+ transactions; trained an optimized XGBoost model with class-imbalance handling, improving precision on high-risk customers by 22% and delivering a production-ready workflow. Developed Web Research Agent, Paper Scraper Agent, and Evaluator Agent workflows to surface answers from user-provided research papers. Led a Face Authentication PoC using a NIR camera for depth imaging to mimic state-of-the-art smartphone unlocking techniques via embedding vectors (PointNet++).

Developed a fraud-detection pipeline on Databricks using SQL for large-scale feature extraction and window-based behavioral signals; performed exploratory profiling on 20M+ transactions and trained an optimized XG Boost model. Conducted multi-agent RAG harnessing Graph for academic research. Built end-to-end fraud-detection workflow and prepared actionable insights for stakeholders.