Data Science Portfolio – Esmaeil Pourjavad

Showcasing professional and self-driven projects in Data Science

View on GitHub

🌍 Seasonal Forecast Models Evaluation Pipeline

Author: Esmaeil Pourjavad
Domain: Data Science | Predictive Modeling | Satellite Data


📌 Project Overview

This work is part of Sphere, a large European project on water availability and climate prediction.
Here I present only my contribution: the design and implementation of a seasonal forecast evaluation pipeline.

The project delivers an end-to-end data science workflow for evaluating and benchmarking predictive models at scale.
Although applied to seasonal climate forecasts (Copernicus C3S multi-model datasets, 1993–2015), the pipeline’s design, tooling, and methodology are generalizable to domains like finance, insurance, and risk analytics.

The workflow operationalizes the full lifecycle:
data ingestion → preprocessing → feature engineering → model evaluation → visualization & reporting,
providing reproducible, automated, and scalable analytics for decision-making under uncertainty.


⚙️ Pipeline Components

1️⃣ Data Acquisition & Orchestration

Keywords: ETL pipelines · workflow orchestration · API integration · data validation


2️⃣ Data Preprocessing & Feature Engineering

Keywords: data wrangling · anomaly detection


3️⃣ Model Evaluation Framework

Framework implemented in Python (xarray, NumPy), R for multi-model benchmarking.

Keywords: benchmarking · probabilistic forecasting


4️⃣ Visualization & Reporting

Keywords: geospatial visualization · statistical dashboards


📊 Tech Stack


📊 Example Results

Some outputs from the evaluation pipeline:

Spatial Map
Spatial Map

Heatmap
Heatmap

Boxplot
Boxplot


📂 Code Snippets

(Full pipeline code is maintained privately due to project restrictions; here are representative excerpts.)


🔑 Data Science Relevance

Applicable beyond climate science → insurance, finance, risk scoring, fraud/anomaly detection.


🎤 Conferences & Publications