Mohamed Ben Hmida

– Development of AI solutions for competence evaluation and workplace safety. – Design and implementation of RAG pipelines using LangChain, embeddings and vector search (Qdrant). – Development of REST APIs with FastAPI to integrate AI services into internal platforms. – Optimization of semantic search pipelines to improve LLM response relevance. – Development of Computer Vision prototypes for fall detection and workplace risk detection. – Mentoring and technical support for AI and Data Science PFE student projects.

– Led the development of an AI-powered ergonomic posture monitoring system, contributing to A.F.In's innovation roadmap in workplace safety. – Built a full Computer Vision pipeline (YOLOv8 + MediaPipe + Metrabs 3D) improving posture detection accuracy by 35%. – Developed a custom ANN model for posture risk classification, achieving 92-95% accuracy. – Implemented ISO-based ergonomic angle computation (neck, trunk, shoulder, elbow) with 95% consistency across frames. – Designed and deployed a real-time monitoring dashboard (Streamlit), improving ergonomic risk visibility for supervisors by 60%. – Optimized real-time inference pipelines, increasing processing speed by 40% through efficient preprocessing and EMA smoothing. – Implemented a robust multi-camera tracking system, reducing ID-switching errors by 70%. – Built analytics tools to visualize angle evolution, flags, alerts, and performance trends, reducing supervisor review time by 50%. – Contributed to the physical & logical architecture for an upcoming industrial deployment at client sites. – Collaborated with AI and software teams to integrate the module into A.F.In's internal safety platform. – Documented experiments, architectures, testing procedures, and delivered technical reports for internal use.

– Designed and created an Intelligent Trading Platform named iTrade which helps customers to make investment decisions. – Used predictive models like LSTM to predict the future prices of assets. – Applied portfolio optimization techniques to ensure portfolio diversification while having a low risk. – Participated in the process of data cleaning and data integration. – Developed a sentimental analysis model to predict the Tunisian stock market movements based on the news. – Built different investment strategies to maximize the return and lower the risk for the customers. – Deployed the model into a web application using Flask.