Transforming BI with Pytorch: Deep Learning Solutions for Customer Segmentation

Welcome to our next Python Meetup! Whether you're a seasoned developer or just beginning your Python journey, this event is designed to help you dive deeper into the language, meet fellow enthusiasts, and share insights. We’ve lined up an exciting main technical talk that will showcase innovative ideas in Python, followed by lightning talks where community members present quick, interesting projects or tips. It’s a great opportunity to learn, network, and collaborate.

Feel free to grab some refreshments, get comfortable, and get ready for an engaging evening of Python!

Date: 20 May, 2025
Time: 6pm-8pm
Location: Improving, 10111 Richmond Ave. # 100, Houston, TX 77042

### Welcome and Networking

• Time: 6:00 PM – 6:30 PM
• Description: Attendees check in, grab refreshments, and network with fellow Python enthusiasts.

### Opening Remarks

• Time: 6:30 PM – 6:40 PM
• Speaker: Dillon Niederhut
• Description: Brief introduction to the event, agenda rundown, and announcements.

### Main Technical Talk

• Time: 6:40 PM – 7:20 PM
• Title: Transforming Business Intelligence with Pytorch: Deep Learning Solutions for Customer Segmentation
• Speaker: Fabrizio Chavez
• Description: As of late, neural networks and large foundational models have been applied to many domains. Yet, one area where these architectures have not outperformed older shallow learning methods has been in the tabular data domain. One common business application that leverages tabular data is customer segmentation, a necessary process for targeted marketing and product development. Traditionally, industry professionals have use qualitative methods to solve this problem. However, within the last 15 years, surveys and clustering algorithms have started to gain traction. Despite the rise of these practices, research and application of these algorithms has been restricted to shallow machine learning methods, without empirical proof on the overall accuracy. This is far from optimal since the clustering problem contains no labels, meaning practitioners need to subjectively interpret their results. This presentation introduces a deep-learning architecture, developed using PyTorch, that was specifically designed for clustering tabular data. Benchmarking against various tabular datasets demonstrates that this method performs best overall and has the potential for providing explainable results.
• Bio: My name is Fabrizio Chavez (Fab), and I’m a last year Ph.D. in the Human Computer Interaction (HCI) program at Rice University. Both my research and industry experience deal with quantitative methods for analyzing user and product data at scale. For context, my projects go from computational models that mimic human behavior to quantitatively analyzing chatbot logs to data-mine usability issues with the product. When it comes to Python, I primarily use it for Deep Learning and NLP tasks, otherwise you can find me in R doing statistics. My latest obsession has been applying neuro-symbolic systems to tabular data

### Lightning Talks

• Time: 7:30 PM – 8:00 PM
• Description: 5-minute talks on various Python-related topics. Lightning talks are meant to be quick and informal, showcasing small projects, tools, or quick tips. Lightning talk sign-ups are at the event!
• Format: Each speaker has 5 minutes to present, followed by 2 minutes for questions.