Franco – Python, Data Science, Machine learning

The company, a manufacturer in the US, specializes in crafting bespoke software solutions that elevate online systems. Their primary goals are to enhance security measures, optimize expenditure, foster seamless collaboration, and empower businesses with profound data-driven insights. With millions of rows of data available at the store-product-week level, this project aimed to create a price elasticity model using mixed models to retrieve coefficients. Additionally, a streamlit app was provided to showcase the model's results and enable re-training as needed.

Franco carried out the following responsibilities:

• performed data exploration and understanding of the client database;
• developed full-cycle of the model from scratch;
• interpretated and evaluated the model;
• developed streamlit app to show results and retrigger training easily;

The project involved conducting demand forecasts for various retailers and manufacturers. This was achieved by employing a combination of time series analysis and machine learning techniques, specifically leveraging XGBoost, while considering multiple features for enhanced accuracy.

Among others, Franco carried out the following:

• explored and comprehended the client's database thoroughly;
• built the model from scratch, covering all development stages;
• interpreted and assessed the model's performance comprehensively.

The client is the largest marketplace for booking accommodation and flights in Latin America. The goal of the project was to optimize the list of hotels that are shown to the client when looking for a specific city. The revenue was increased by 3-7%, depending on the city.

Franco accomplished the following tasks:

• created a training model pipeline;
• engaged in feature engineering, integrating intricate features like density-based clustering for locations;
• assessed the model's performance;
• executed AB testing;
• collaborated with the MLOps team for deployment.

The project focused on tracking media ratings and store-level sales, aiming to develop regularized models for understanding the impact of media campaigns on clients' sales. The primary objective was to construct a cohesive model that evaluated each campaign's effectiveness based on coefficients and ROI. Additionally, the model offered insights for budget optimization in the subsequent year, leveraging saturation curves to pinpoint growth opportunities for individual campaigns.

Franco successfully completed the following tasks:

• conducted data exploration and gained an understanding of the client's data;
• developed and trained various models;
• interpreted and evaluated the models;
• communicated results to stakeholders and less technical individuals;
• enhanced processes through automation and code implementation;
• managed a team of five technical analysts, ensuring high-quality outputs.