SOUIYAH Hajar

– Designed and implemented a multi-agent architecture for automated data analysis, visualization, and business recommendations. – Integrated database connectivity PostgreSQL and CSV/JSON ingestion with automated data cleaning workflows. – Developed automatic database schema generation and metadata extraction stored in MinIO. – Implemented Elasticsearch indexing for fast retrieval of query results and metadata. – Integrated Mistral LLM via Ollama to generate SQL queries, and suggest relevant visualizations. – Developed a conversational chatbot leveraging RAG to enable users to customize dashboards, query data, and receive natural language explanations and recommendations. – Developed backend with FastAPI and asynchronous task execution using Celery to ensure scalability and responsiveness. – Tech stack: Python, Ollama, Openrouter, Mistral LLM, RAG, FastAPI, Elasticsearch, Minio, Celery, Docker, PostgreSQL, Git.

– Developed a web-based banking alert management application to detect suspicious transactions and prevent fraud. – Integrated Random Forest and Logistic Regression models to improve fraud detection accuracy, reduce false positives, and uncover hidden transaction patterns. – Built a multi-role authentication system (Administrator, Alert Manager, Analyst, Client) with granular permission control and activity history tracking. – Created interactive dashboards and visualization tools to analyze alerts, track status changes, and generate operational insights. – Implemented automated workflows for alert validation, including client verification forms, response history logging, and alert closure processes. – Tech stack: Machine Learning, Django, HTML, CSS, JavaScript, MySQL.