Super Data Science: ML & AI Podcast with Jon Krohn 939: Mixture-of-Experts and State-Space Models on Edge Devices, with Tyler Cox and Shirish Gupta
21 snips
Nov 11, 2025 
Tyler Cox, a distinguished engineer at Dell focusing on on-device AI, and Shirish Gupta, leader of Dell's AI initiatives, dive into the world of state-space models and their efficiency in processing information. They discuss the benefits of running AI workloads locally, including cost savings and improved privacy. The duo also highlights the innovations of the Dell Pro AI Studio and partnerships with IBM for granite models. Listeners will love learning about the Mamba architecture and how it enhances AI performance across devices.
AI Snips
Chapters
Books
Transcript
Episode notes
Unified Toolkit For On‑Device AI
- Dell's AI PC solution (formerly Dell Pro AI Studio) unifies model catalog, on-device framework, and cloud management for enterprise deployment.
- It abstracts hardware and runtime diversity so developers can deploy local AI as easily as cloud APIs.
Use On‑Device AI Deliberately
- Do treat on‑device AI as an option in hybrid architectures to reduce cost, latency, and privacy exposure.
- Evaluate which inferences should run locally versus in cloud based on latency, cost, and data sensitivity.
Leverage Model Management And Tooling
- Use the provided command‑line and model management tools to control versions and deploy consistent bundles across mixed hardware.
- Package models, runtimes, and installers together so IT can push solutions at scale without manual per‑device tuning.
