SOUHAIB GARAAOUCH

Developed and deployed a full-stack Generative AI application to bridge the critical gap between raw vibration data and actionable maintenance recommendations, delivering reliable, low-latency (<5s) action plans for industrial machinery. Key Achievements & Responsibilities: • Engineered a Full-Stack Proof of Concept using FastAPI (Python) for the backend API and Next.js (React/TypeScript) for a real-time, interactive dashboard that visualizes machine data and AI-generated results. • Designed and Implemented a Sophisticated "Controlled Generation" Prompt Pipeline with LangChain, using modular templates and Pydantic schemas to ensure reliable, structured, and safe JSON output from the Large Language Model. • Solved a Critical Performance Bottleneck by migrating the inference engine from standard APIs to the high-speed Groq API. This leveraged specialized LPU hardware and advanced techniques like speculative decoding to reduce recommendation latency by 90% (from over 40s to ~4-5s). • Containerized the Entire Application (backend and frontend) using Docker and Docker Compose, creating a portable, reproducible environment for one-command deployment. • Integrated LangSmith for End-to-End Observability, enabling real-time tracing, logging, and debugging of every LLM call to monitor performance and ensure the reliability of the AI's reasoning process. • Developed a Quantitative Evaluation Framework using Python to benchmark model performance against a ground-truth dataset, establishing a data-driven process for validation and continuous improvement.

Mission: • Developed and integrated an AI-powered analytics module into a Spring Boot supply chain application to provide data-driven insights. • Engineered a data model in Python to analyze key operational metrics, identifying trends in inventory levels and company performance. • Built robust backend APIs with Spring Boot (Java) to process supply chain data and serve analytical results, enabling better strategic decision-making. • Containerized the analytics service using Docker, ensuring a reproducible and isolated environment for the Python data models.

Mission: • Architected and developed a full-stack customer support platform using a microservice-based architecture to ensure scalability and maintainability. • Engineered the backend API services using Spring Boot (Java) and managed data persistence with MongoDB. • Built a dynamic and responsive user interface for the platform using React.js and Node.js. • Collaborated within an Agile team to define feature requirements, solve technical challenges, and deliver the project within the internship timeline.