Mohamed Amine Fakhre-Eddine

– Led end-to-end development of AI-powered applications, designing and implementing LLM-driven chatbots and RAG systems tailored for legal workflows. – Built modular AI services using LangChain, Python, and TypeScript, integrating advanced retrieval pipelines, document parsing, and domain-specific reasoning for law-focused use cases. – Collaborated with software engineering teams to architect scalable backend and frontend solutions, ensuring seamless integration of AI models into production web applications. – Developed and optimized multi-agent and function-calling chatbot systems, enhancing accuracy, UX, and automation across client-facing legal tools. – Implemented vector search and knowledge-base systems using technologies such as ChromaDB, Pinecone, and custom embeddings pipelines to support high-precision legal information retrieval. – Deployed AI services using Docker, orchestrated cloud environments, and maintained CI/CD workflows for reliable, secure, and scalable performance. – Partnered with product and legal teams to translate requirements into AI-driven features, improving internal workflows and enabling data-driven decision-making.

Internal tools: – Developed a multilingual semantic alignment tool for automotive defect validation, automating cross-language consistency checks and generating rule-based labeling reports, reducing manual review errors by 90% and processing time by 99%. – Built an automated data consolidation app for signal analysis in Excel, integrating cleansing, deduplication, and dynamic filtering, saving ~ 10 hours per week across engineering teams. Stellantis + Capgemini Engineering: – Migrated consultant performance dashboards from Excel to Power BI, automating data refresh via SharePoint and creating interactive visuals for First Time Right and On-Time Delivery metrics, improving visibility and reducing reporting effort by 75%.

– Developed a research paper analysis chatbot using LLMs (Hugging Face, LangChain, PyTorch), enabling semantic search and conversational insights for scientific literature. – Deployed on Docker and integrated with ChromaDB and SvelteKit frontend for interactive use.

– As a team of two students, we collaborated with the Mechanical Engineering department to create an RNN-based predictive model for rotating machinery fault detection, improving classification accuracy. – Conducted full-cycle experimentation, data acquisition, visualization, preprocessing, and model validation using Keras, Scikit-learn, and SciPy.