Developing an emotionally intelligent conversational AI assistant inspired by the movie 'Her', with dual processing options: fully local execution for maximum privacy or integration with ChatGPT Realtime API for enhanced performance and lower latency. Features natural emotion detection and expression in conversations.

Built a multi-agent AI research assistant using CrewAI and LangChain, with dynamic model switching between local Deepseek R1 via Ollama and OpenAI model o1 through their API, guiding users through automated search, analysis, and structured report generation via a Streamlit interface.

Developed a Streamlit-based RAG application that ingests healthcare insurance PDFs, chunks and embeds text with FAISS + AWS S3, and delivers context-aware GenAI answers via Claude/Amazon Titan.

Produced a predictive model for power consumption in Tetouan City using Machine Learning in Python, leveraging weather and time-related variables such as diffuse flow from UCIML Repo.