#foundationmodels #ai #datascience Jazmia Henry joins @JonKrohnLearns to break down what it actually takes to build end-to-end foundation models for the energy industry. From wrangling decades of handwritten oil-and-gas documents into usable training data, to bespoke tokenizers, reinforcement learning, and inference at scale, Jazmia walks through every stage of the stack. Along the way she explains why reinforcement learning models are "bursty," what reward hacking is and how her Grounded Continuous Evaluation framework fixes it, and revisits the 2023 NeurIPS paper that argued, to widespread skepticism at the time, that scaling bad data degrades model performance. This episode is brought to you by: • Acceldata: https://acceldata.io/ Interested in sponsoring a SuperDataScience Podcast episode? Email natalie@superdatascience.com for sponsorship information. In this episode you will learn: • (00:00:00) Introduction • (00:03:25) Why oil and gas needs AI • (00:12:19) The four stages of foundation model building • (00:20:13) Why reinforcement learning models are "bursty" • (00:34:02) Reward hacking and the GCE framework • (00:48:12) Why scaling bad data makes models worse Additional materials: https://www.superdatascience.com/995