Even the Chip Makers Are Making LLMs
In this episode, NVIDIA VP of Generative AI Keri Britsky explains why a GPU chip maker is now deeply involved in building large language models (LLMs). She describes NVIDIA’s extreme hardware‑software co‑design process, where model development informs GPU architecture, precision formats (FP8, FP4) and new engines like the Context Memory Engine. Britsky also details the NemoTron family of models, hybrid transformer‑Mamba architectures, and disaggregated inference frameworks that improve token efficiency and memory usage. The discussion highlights how these innovations enable more scalable, accurate AI systems across vision, speech, and language tasks.
AI-Assisted Coding Needs More than Vibes; It Needs Containers and Sandboxes
In this episode, Docker President Mark Cavett discusses how containers are becoming essential for safely running AI‑generated code, emphasizing the need for hardened images to bridge the trust gap. He explains Docker’s new open‑source Docker Hardened Images (DHI) catalog, which...
No Need for Ctrl+C when You Have MCP
In this episode, Ryan Donovan interviews David Soria Parra, co‑creator of the Model Context Protocol (MCP) and a technical staff member at Anthropic. They discuss the origin of MCP as a solution to the copy‑paste friction when using LLMs, its evolution...
Why Stack Overflow and Cloudflare Launched a Pay-per-Crawl Model
In this episode, Stack Overflow’s Janice Manningham and Josh Zhang chat with Cloudflare VP Will Allen about the newly launched pay‑per‑crawl model that lets publishers charge crawlers for access. They explain how AI‑driven content scraping has upended the traditional open‑versus‑block...
Data Is the New Oil, and Your Database Is the only Way to Extract It
In this episode, Ryan interviews Shireesh Thota, Corporate Vice President of Azure Databases at Microsoft, about the rapid evolution of Microsoft's database offerings, including SQL Server, Cosmos DB, and Postgres, and how they fit into a unified Azure data platform....
Even Your Voice Is a Data Problem
In this episode, Ryan interviews Scott Stephenson, CEO and co‑founder of Deepgram, about the latest advances in voice AI, focusing on how deep learning improves speech‑to‑text and text‑to‑speech accuracy across diverse dialects and noisy environments. They discuss Deepgram’s scalable, affordable...