Yassine Ben Abdelaziz

LLM Prompt Management System • Engineered and deployed a prompt management tool for AI developers, addressing the needs for prompt versioning, experiment reproducibility, evaluation, and collaboration throughout the prompt lifecycle. • Integrated GitHub to version prompts, track the history of changes, and ensure a single source of truth for prompt, models and configurations. • Supported running experiments across multiple models from providers such as OpenAI as well as open-source models from Hugging Face, leveraging Semantic Kernel for orchestration. • Delivered an evaluation framework that supports both standard and user- defined evaluations on datasets, using RabbitMQ to asynchronously process evaluation workloads and MinIO for dataset storage. • Authored a Python SDK and CLI to deliver a developer-first experience, enabling seamless integration into existing codebases and CI/CD workflows, along with comprehensive documentation. Technologies: NextJs, Dotnet, FastAPI, Postgres, RabbitMQ, MinIO, LLM, Semantic Kernel, GitHub API, Auth0, SignalR, Clean Architecture, CQRS, DDD, BDD, xUnit, Pytest

Backend Optimization and Feature Enhancement for Residential Incident Management and Resolution Platform. Tasks : - Collaborated in the development of features to enhance communication between residents and property managers, improving incident reporting and request tracking via QR codes. - Implemented improvements to the platform's notification system, enabling automatic status updates and multi-modal notifications (email and SMS). - Redesigned backend architecture, transitioning from a Strapi headless CMS to a more flexible and scalable environment, solving issues like duplicated data and inadequate database design. - Enhanced user experience (UI/UX) by addressing key front-end design issues, improving interface intuitiveness and consistency.

Digitalization of Cancer Treatment in Tunisian Hospitals. - Revolutionized patient data management by developing a digital platform that replaced traditional paper records and improved data accessibility. - Built a core system around fully customizable and reusable dynamic forms, empowering oncologists to tailor records to their specific needs and effectively manage all types of cancer. - Fostered collaboration and standardized treatment protocols by implementing a template-sharing feature. - Focused on user-friendliness through intuitive design, ensuring a seamless workflow for medical professionals.

Design and development of a decision-making mobile application. Tasks : - Created an ETL script to extract data from a closed UNIX environment and integrate it into a relational database. - Implemented a dashboard to monitor company activity and track important performance indicators. - Designed graphical tools for periodic visualization of the commercial situation summary.