Mohamed Hassani

Led end-to-end development of an AI-powered image generation platform for client exhibitions. Set up and deployed workstations to run image-generation models; built Python applications with and without GUI to let users choose artistic styles and generate personalized images from photos. Integrated gesture control and touchscreen capabilities for triggering photo capture. Deployed a server to save and deliver outputs, collect statistics, manage user data, and generate QR codes for easy downloads. Customized outputs with styles and visual effects to meet exhibition requirements, and ensured system scalability to handle high user volumes. Conducted testing and monitoring to maintain reliability.

Designed and implemented Reinforcement Learning from Human Feedback (RLHF) workflows to enhance large language models' code generation and function-calling capabilities. Created challenging programming projects across Python, C++, Swift, JavaScript/HTML, and Java; developed comprehensive coding challenges for code generation, bug fixing, refactoring, and solution reasoning to strengthen LLM capabilities. Served as technical reviewer and QA, evaluating AI-generated code and contributor work to ensure high-quality training datasets. Implemented accurate code solutions for complex simulated user requests and optimized tool selection and function calling. Curated high-quality training data following strict linguistic, formatting, and localization guidelines for improved generative AI performance in remote settings.

Specialized in data annotation and quality assurance for AI language models. Created high-quality training data and reviewed agents' outputs to identify inaccuracies, improve error rates, and ensure relevance. Adhered to detailed annotation guidelines and maintained confidentiality of proprietary data and models. Maximized usefulness of contributions by aligning work with project goals and enhancing overall model quality.