Love Data Week 25 Lightning Talks: AI & Data

Speakers: Kaze W. K. Wong, Doran Larson, and Brian Wingenroth

Data pipeline for athletic performance optimization at JHU (Wong)

Machine learning and AI tools have made tremendous progress in computer vision tasks such as image segmentation, action localization, and pose estimation. These are tasks that can be used to extract useful analytics out of videos of athletes performance, hence can be used gain insight about training methods and help optimize performance. However, most of existing AI tools stop at the level of research code instead of being a production-ready tool the users can just deploy. In this talk, I am going to talk about a research and engineering effort I lead to build a production-ready data pipeline for the Johns Hopkins Track and Field team, specifically for the high jump event. I will discuss the prospects and challenges to building such an end-to-end pipeline that not only is going to result in research papers in machine learning, but translates to improvement in the performance of athletes. I will discuss these challenges from different prespectives, including an AI researcher, a software engineer, a coach, and finally an athlete. Specifically, we are going to go from data collection to integrating different continuous integration/continuous deployment (CI/CD) and MLOps tools such that our machine learning pipeline can run whenever new data is generated.

Archiving Prison Witness: History, Power, and Entitling (Larson)

This talk will review the inception and history of the APWA, discuss the potential power of first-person prison witness to bend the arc of carceral history, and discuss APWA policies and practices that seek to make platforming of prison witness non-extractive by granting rights and ownership to its 1000+ (and growing) authors.

The OIDA Toolbox: Supporting Research on the Opioid Industry Documents Archive Across Methods and Disciplines (Wingenroth)

The OIDA team at JHU has developed and released the OIDA Toolbox, a suite of tools supporting diverse approaches to conducting research with the archive. OIDA, the UCSF-JHU Opioid Industry Documents Archive, houses over 4 million documents from the opioid industry. This presentation will demonstrate how researchers across disciplines can access and analyze these materials through multiple pathways - from traditional search and download to cloud-based computing environments - designed to support users at all levels of technical expertise.

Registration link: https://bit.ly/JH_lovedataweek

** This session is part of Love Data Week 2025. To attend this session, first register here: https://bit.ly/JH_lovedataweek and then follow the instruction under "Registration and Creating an Itinerary" in the description to add this session to your itinerary. **