Papers to Podcasts Unplugged transforms peer-reviewed research papers into engaging audio stories designed for a broader audience. Through this initiative, we reimagine science communication—bridging the gap between dense academic literature and curious minds everywhere. Each episode is generated using AI tools and narrated in plain language to preserve the integrity of the science while making it more approachable. Check out some of our transformed research papers below.

Mining the Future: The Promise and Challenge of Direct Lithium Extraction

---

This episode is a deep dive on a recent ⁠review article⁠ published in the Chemical Engineering Journal and explores the emerging world of Direct Lithium Extraction (DLE) and its potential to revolutionize the way we power our electric future. As demand for lithium-ion batteries skyrockets, traditional lithium sources, like salt flat evaporation ponds, are struggling to keep up. We'll examine the challenges posed by traditional methods, including their environmental impact, high energy consumption, and limited scalability. 

Decoding Complexity: Grouping Chemical Mixtures with Machine Learning & Analytical Chemistry Fingerprints

---

This episode takes a deep dive into a ⁠2019 study⁠ published in the PlosOne Journal that uses machine learning to understand and group complex chemical mixtures. We explore how these techniques can analyze large analytical chemistry datasets, offering new ways to categorize substances based on their chemical fingerprints. From GC-MS to ToxPi, and from confusion matrices to supervised and unsupervised learning, we break down the science behind our approach. Join us as we discuss the challenges of chemical mixture analysis and how machine learning can help make sense of complex environmental data.

Listen on Spotify.

Finding the Needle in a Haystack: Optimization in Highly Constrained Systems

---

This episode takes a deep dive into a recent ⁠Industrial & Engineering Chemistry Research study⁠ that explores how machine learning can tackle some of the hardest optimization problems in energy systems—those with hundreds or even thousands of constraints. We unpack how surrogate models and classification-based constraint handling are used to optimize waterflooding operations in oil reservoirs, from benchmark models like Egg to large-scale, real-world systems like UNISIM. Along the way, we break down concepts like neural network surrogates, feasible vs. infeasible regions, and why traditional constraint-handling methods often fall short at scale. Join us as we discuss how combining regression, classification, and data-driven optimization can dramatically reduce computational cost while still delivering economically meaningful and physically feasible solutions in complex engineering systems.