I'm Mohammadali Azani

Developed for the World Health Organization (WHO), this mobile & web application enables rapid structural assessments of healthcare facilities during Ebola and Marburg outbreaks. Built with Flutter, Dart, and Clean Architecture, the tool supports offline assessments, real-time progress tracking, dynamic scoring, and multi-format exports (PDF, Excel, Word).

LLM-MathEval is a modular, Dockerized pipeline for fine-tuning, evaluating and benchmarking LLMs on mathematical reasoning: adapter-based tuning on the MathChat benchmark; automated JMeter load tests with CPU/GPU monitoring and performance-load correlations; end-to-end QA accuracy pipelines outputting CSV/JSON; Phoenix observability with carbon-footprint tracking; and fully reproducible deployment.

A full-stack web application creating a comprehensive yoga platform. The frontend delivers a fully responsive UI with SSR for optimal SEO performance using Nuxt 3, Vue 3, and Tailwind CSS. The backend provides RESTful APIs managing yoga-related data, teacher profiles, and activities, built with Express.js and PostgreSQL. Both components are deployed on Vercel.

Developed a Django-based backend for a vehicle management mobile app featuring real-time 3D visualisation. Implemented RESTful APIs supporting component-level access control, user authentication, and vehicle monitoring. Built with Django, PostgreSQL, JWT, and Guardian, deployed on Azure.

This project focuses on the classification of blood cell images into eight distinct classes using advanced deep-learning techniques. Leveraging state-of-the-art convolutional neural network architectures, the goal is to achieve high accuracy in distinguishing between different types of blood cells, which is crucial for medical diagnostics and research.

This project designs a strategic e-bike charging infrastructure along linear cycle paths like the Danube. Using a graph-based approach and a Mixed-Integer Programming (MIP) model, it minimises the maximum distance between charging stations while adhering to budget constraints.