Mirgafar Mammadli

Predict-Loan-Default-Risk-for-Farmers-Using-Geospatial-Data

June 2, 2026 – Present

Combine tabular data on agro loans with geospatial indicators (e.g., drought exposure, proximity to markets) to model default risk in Azerbaijan.

Segment-Customers-of-Azerbaijani-Retail-Chains-Using-Clustering

May 24, 2026 – Present

Use unsupervised learning to segment customers based on purchasing behavior to help local chains (e.g. Bravo, Araz) optimize marketing.

Detect-Crop-Diseases-Using-Aerial-Images

May 18, 2026 – Present

Train a deep learning model to classify plant diseases using high-resolution drone images, supporting precision agriculture in Azerbaijan.

Build-and-Query-a-Feature-Store-for-ML-Training-Serving

May 12, 2026 – Present

Create a reusable feature store to serve features for both training and inference using a modern tool like Feast. **Requirements:** - Define entities and feature sets (e.g., customer, transaction) - Use Feast for feature store setup - Materialize features from batch + stream sources - Query feature vectors for online/offline use

Estimate-the-Causal-Impact-of-Marketing-Campaigns

May 4, 2026 – Present

A causal inference pipeline for measuring true marketing campaign lift — DiD, CausalImpact, EconML meta-learners, and cluster bootstrap in one notebook.

Build-a-Multi-Modal-Product-Classifier-Using-Text-and-Images

April 26, 2026 – Present

Classifying fashion products using both images and text — because a picture of a shirt and the words "men's casual shirt" together tell you more than either one alone. Built with ResNet-18 + DistilBERT, fused through a custom gated cross-attention mechanism.

Combine-Prophet-and-XGBoost-for-Time-Series-Forecasting

April 18, 2026 – Present

Combine-Prophet-and-XGBoost-for-Time-Series-Forecasting — GitHub repository

-Hospital-Readmission-Prediction-with-Explainable-AI-XAI-

April 9, 2026 – Present

focuses on predicting the 30-day readmission risk for diabetic patients using a dataset of 70,000+ clinical records from US hospitals. The primary goal is to provide a high-performing predictive model while ensuring transparency through Explainable AI (XAI) techniques.

migafar

May 31, 2025 – Present

migafar — GitHub repository

Real-Estate-Price-Prediction-in-Baku

May 28, 2025 – May 28, 2025

> Baku Apartment Price Prediction > This project aims to predict rental apartment prices by extracting data from bina.az. > Technologies Used: > - Web Scraping: requests, BeautifulSoup > - Data Processing: pandas, numpy > - Machine Learning: scikit-learn