Hamidreza Zoraghein

Data Scientist focused on predictive modeling, LLMs, geospatial analytics, and intelligent decision systems.

Data Scientist · Predictive Modeling · LLMs · Energy & Climate Systems

I design predictive models and simulation-based analytics to support long-term planning and decision-making in energy, transportation, public health, and climate-related systems.

I also generate AI-enabled pipelines based on Large Language Models to generate insights and improve evidence-based understanding.

With over 10 years of experience, my work focuses on combining machine learning, statistical modeling, and spatial analysis to understand system behavior under uncertainty and future scenarios.

🔍 Core Areas of Focus

  • Predictive modeling & forecasting using machine learning and simulation
  • LLMs using Retrieval-Augmented Generation (RAG), fine-tuning, and prompt engineering
  • Scenario analysis for energy demand, electrification, and infrastructure planning
  • Geospatial modeling across large spatial and temporal scales
  • Uncertainty-aware analytics to support real-world decisions

🚀 Selected Work

Northeast Freight Corridor Charging Plan
Predictive and simulation-based modeling to support proactive grid planning under fleet electrification and future demand scenarios for utilities in the Northeast U.S.
→ View Project

NYC Climate Vulnerability Modeling
Spatial deep learning and simulation frameworks for high-resolution population vulnerability analysis; accepted for publication in PNAS.
→ View Project

EVRAG — Language Language Models for Model Interpretation (Side Project)
A Retrieval-Augmented Generation pipeline designed to help interpret and communicate results from technical EV charging reports. Language models are used as a supporting layer, not a replacement for quantitative modeling.
→ View Project

AI-enabled COVID-19 forecasting
End-to-end machine learning workflow using XGBoost and LSTM to generate weekly forecasts of COVID-19 incidence rates at U.S. county level. → View Project

🧠 Modeling Philosophy

My approach emphasizes understanding systems, not just fitting models. I prioritize:

  • interpretability and transparency over black-box accuracy
  • simulation and scenario exploration alongside machine learning
  • models that inform planning decisions, not just predictions

Language models are used selectively to augment analysis and interpretation when working with complex documentation and unstructured information.

📌 Highlights

  • 10+ years of applied experience in predictive modeling and simulation
  • Published research in Proceedings of the National Academy of Sciences (PNAS) and Nature Communications
  • Work spanning energy systems, climate impacts, population dynamics, and public health

📬 Get in Touch