Xalid Ahmadov

Fully automated soil moisture prediction platform for Azerbaijan's diverse climate regions using Python ETL pipelines, DuckDB Medallion Architecture, region-specific XGBoost models, GitHub Actions CI/CD, a React dashboard, and smart irrigation email alerts.

Innovative solution to make urban traffic more efficient and sustainable, leveraging AI to optimize traffic flow and reduce congestion across city networks.

Time-series forecasting on historical Bitcoin data using a Dense Neural Network with windowed sequences. Evaluated with MASE, MAE, RMSE, and MAPE against a naive baseline.

Retrieval-Augmented Generation system for answering questions from .docx documents using FAISS vector search and SentenceTransformer embeddings.

Local LLM pipeline using LangChain with Phi-3-mini (GGUF) to build multi-stage text-generation chains, entirely without cloud APIs.

Text emotion classifier using spaCy preprocessing, TF-IDF vectorization, and a Random Forest model to distinguish emotions such as joy, fear, and anger.

Text-generation and chain-of-thought reasoning workflows using Phi-3-mini-4k-instruct with structured prompts and multi-step instruction-following tasks.

Transfer-learning image classifier for rice variety recognition, fine-tuning EfficientNetB0 on a custom dataset with Keras data augmentation layers.

Binary image classifier built with a custom CNN in Keras, trained on a structured cat-and-dog image dataset with batch preprocessing and Adam optimization.