Elham Afzali

University of Manitoba

Contributed original research at the intersection of statistical theory and machine learning, producing peer-reviewed publications and practical tools for model evaluation in complex, high-dimensional data settings.

• Designed MS-KSD-Bayes, a density-weighted extension of KSD-Bayes that eliminates mode-proportion blindness in score-based inference while preserving posterior consistency and asymptotic normality. The new weighting scheme restores correct mode representation and markedly improves Bayesian reliability in complex multimodal settings.
• Applied MS-KSD-Bayes to synthetic benchmarks and two real-world datasets—Galaxy velocity measurements and the lung-cancer gene-expression panel (GEO GSE68571)—demonstrating substantial accuracy gains over existing KSD-based methods in multimodal and heavy-tailed settings without compromising unimodal performance, and released an open-source Python library implementing weighted Stein-discrepancy computations to ensure full reproducibility.
• Developed GF-KCSD, a gradient-free conditional goodness-of-fit test in RKHS, delivering theoretically grounded diagnostics when score-based methods are hindered by intractable gradients; validated its accuracy through comprehensive simulation studies and comparisons with gradient-based tests.
• Enhanced GF-KCSD with bootstrap-calibrated critical values and a detailed measurement-error analysis, boosting interpretability and α-control. Additional simulations and a high-dimensional brain-MRI age-prediction study confirmed stable type-I error, strong power, and a three-fold speed-up over gradient-based KCSD.
• Led the design and implementation of a hybrid VAR-LSTM deep learning model to forecast COVID-19 dynamics in Canada. By integrating classical time-series analysis (VAR) with neural networks (LSTM), the model achieved 93.07% accuracy on long-term predictions, substantially outperforming standard models.
• Proven ability to manage multiple high-level responsibilities by independently conducting advanced methodological research, developing and evaluating statistical algorithms, and completing publication-ready work — all while maintaining full-time professional commitments and consistently meeting project deadlines.

Lockport Archaeology Project

• Served as lead analyst in an interdisciplinary project exploring historical narratives of the Lockport region, coordinating collaboration between archaeology and statistics teams.
• Curated, cleaned, and validated large-scale archaeological datasets, ensuring data integrity and readiness for rigorous quantitative analysis.
• Applied advanced statistical techniques and high-dimensional visualization to uncover meaningful patterns and temporal trends in complex artifact and excavation records.
• Produced insights that bridged statistical rigor with archaeological interpretation, demonstrating the power of quantitative methods to illuminate cultural and historical contexts.

Refah Bank

• Led the processing and analysis of large-scale banking data to inform executive decision-making.
• Directed the development of interactive dashboards and daily reports in Power BI and Excel, delivering critical insights on schedule.
• Oversaw customer-analytics and data-mining initiatives using Python and SQL, guiding analysts to actionable findings.
• Architected and managed a new data-warehouse solution and end-to-end ETL pipelines (SSIS), establishing the infrastructure for enterprise business intelligence.

• Reviewed academic papers and graduate theses for internal research validation and knowledge dissemination.

• Automated end-to-end HR reporting workflows, replacing manual spreadsheets with scheduled ETL jobs and interactive Power BI reports.
• Designed and maintained real-time dashboards visualizing headcount, turnover, time-to-hire, and compensation metrics.
• Provided ad-hoc analyses on workforce demographics, engagement surveys, and overtime costs to support strategic planning initiatives.

• Developed statistical risk assessments and predictive models for financial forecasting and credit-risk analysis.

Entekhab Factory

• Assisted in collecting and organizing customer data to support market-research and brand-development strategies.
• Collaborated with marketing and analytics teams to evaluate promotional-campaign outcomes and deliver data-driven insights that enhanced customer engagement.
• Supported reporting and visualization of customer trends using Excel and basic statistical tools.