An end-to-end tabular machine learning pipeline predicting accommodation demand from historical reservation searches. Built with Python, Pandas, Scikit-learn, LightGBM, and XGBoost. Features strict out-of-time temporal cross-validation, Jalali calendar holiday proximity encoding, and empirical-Bayes accumulation kinetics, achieving a 16.36% Pooled WAPE with the champion model compared to a 23.05% baseline across 321 destination cities.
GitHub Repository ContextNARGES KHALEQI // SPACE FLIGHT JOURNEY
ORBITAL LAUNCH // TRAJECTORY ARMED
·····
[STATION 01]ML Research Observatory
[STATION 02]Game Projects Arcade Outpost
[STATION 03]Communications & Mobile Engineering Command
[AI CORE]Neural AI Core & Automation Gateway
[DESTINATION]Cyberpunk Megacity Port & Comms
NK-01 GLIDE
SCROLL DOWN TO INITIATE DESCENT TRAJECTORY
Narges Khaleqi — Computer Science Student & Software/ML Builder
Computer Science student at Islamic Azad University (Tehran Markazi). Creator of Haj Mirza (offline Persian word-finding app on Cafe Bazaar), builder of time-series demand forecasting machine learning pipelines (Pol 4 / Jabama competition), and developer of interactive full-stack and AI prototypes.
Featured Engineering Project: POL-4 / Jabama Demand Forecasting (2025)
Other Shipped Work
Haj Mirza — Persian Word-Finding Android App (2026)
Production Android application published live on Cafe Bazaar. Shipped solo in 16 days using React Native, Expo, and TypeScript, featuring offline anagram solving, Persian character normalization, and pre-indexed dictionary datasets.
Cafe Bazaar LinkAdPerk — Telegram Stars Digital Commerce Platform Prototype (2026)
Conversational digital goods bot prototype handling native Telegram Stars payments with webhook signature validation and Firebase Firestore persistence.