learn & Build
nlp in the Age of LLMs
Generative machine learning, specifically in the form of large language models (LLMs), has radically changed the natural language processing (NLP) landscape. Over the last three years, we have seen LLMs produce amazing and, sometimes, interesting (?) results. Today, we will explore the strengths, but more importantly the limitations of LLMs. What is the role of traditional approaches NLP landscape? Are task-specific machine learning models dated? Spoiler. No.
Building Machine Learning Pipelines for institutions
How do we design quality machine learning systems that can reliably process archival records? What happens when those archives contain highly sensitive data? These questions shall frame our journey. This talk is about the lessons I’ve learned from building machine learning pipelines to run across sensitive data. We’ll discuss ethics and its role in an ever-changing AI landscape.
Interrogating Digital Justice as Disaster Recovery
Featuring three project teams across ACLS’s Digital Justice Grant Program and its Digital Extension Program, this panel will interrogate the multifaceted and multivalent relationship between digital justice and disaster recovery within various geographic locales across the Global North and South.
I will be speaking about my work on the ACLS-funded project Personal Writes the Political. We are working with a sensitive collection of 100,000 letters written by predominantly rural women in South Africa during Apartheid. We will be exploring the impact of such a collection and the ethics behind applying machine learning to cultural heritage collections.