A repository to save my small Data Science and AI projects, including AI Engineering applications, data analisys from Kaggle datasets and building of ML and DL models. The projects are developed in my learning process and the predictions and analisys should not be fully considered.

Structure:

• Project 1 (April 2024): Universities Salaries Analisys - My first notebook from Kaggle, using a dataset of universities salaries. I've applied what I was currently learning on data science, so I created some graphs with Seaborn and did some basic analisys with Pandas;
• Project 2 (November 2024): Economic Situation Graphs - Simple application using streamlit, that shows some charts about the economic situation of each country of the world. In this project I learned the basics of streamlit and plotly;
• Project 3 (February-March 2025): Workout Dataset Analisys and Prediction Model - An application using Pandas, Scikit-learn, Statsmodels and Ploty that analyses the correlation between biological measures and the amount of calories burnt during a workout, to create a machine learning model able to predict it. I also created a Streamlit basic application where you can have a overview of the dataset and input values for the prediction model.
• Project 4 (February 2025): Multimodal AI App - An AI application that creates an audio story from an image, based on HuggingFace AI models and UI made with streamlit. I learned the concepts about this project following a Youtube tutorial.
• Project 5 (March 2025): Chatbot - A ChatGPT-like application created with Streamlit and integrating AI models from HuggingFace via TogetherAI. The chatbot suports custom instructions, styles and model choosing. In this project I learned a lot about interfaces with Streamlit and how to use context with LLMs.
• Project 6 (September 2025): AI Agent similar to Perplexity - Learned how to create graphs on LangGraph and use multiple local models to divide tasks and generate structured answers with web searching. Used Tavily for web searching and OLlama for local models. Models utilized: DeepSeek R1:7b for reasoning tasks and Qwen 2.5:3b for answering small queries.