Turning complex data into analytics systems and agentic AI research.
Mahyar Arani works across analytics engineering, data science, business intelligence, geospatial analytics, and agentic AI evaluation. The work combines data modeling, reproducible experiments, and decision-support systems for operational, financial, and research problems.
Professional focus
This work sits at the intersection of analytics, data science, and agentic AI research. Across the portfolio, the recurring pattern is translating messy operational or research problems into structured data products, evaluation pipelines, and decision systems.
The GitHub portfolio adds supporting technical material from public repositories and selected private research systems summarized without exposing internal files.
Core strengths
- Agentic AI benchmark design for temporal leakage, replay, and financial-risk evaluation.
- Analytics engineering with strong attention to data quality, semantic modeling, and exception handling.
- Machine learning and forecasting projects for classification, regression, credit risk, and time series.
- Geospatial analytics and routing-oriented problem solving using ArcGIS and related systems.
- Research-driven thinking across financial risk, sustainability, organizational behavior, and consumer behavior.
Cloud, data, and analytics credentials
Decision-ready data products
The project work focuses on dashboards, data pipelines, model comparisons, geospatial analysis, and reporting systems designed for repeated operational use.
Benchmark-driven research
The current research portfolio studies temporal leakage, counterfactual replay, financial-agent behavior, and systemic risk in AI-mediated decision environments.
Modeling with context
The modeling work emphasizes time-aware validation, feature design, interpretability, and the connection between model output and real decisions.
For project discussions, research conversations, or collaboration opportunities, the best contact point is info@mahyar-arani.com. You can also connect on LinkedIn.