Listening guide
Tobias Macey's AI Engineering Podcast is a technical interview show about how AI systems get built; recent episodes run about an hour, usually with one guest. Here are six of its episodes, each with a Chain of Thought episode to play next.
I host Chain of Thought. We have no affiliation with the AI Engineering Podcast: no partnership, no arrangement, nothing either of us gets from this page.
The AI Engineering Podcast describes itself as a guide to building scalable and maintainable AI systems: architecting AI applications, applying AI at work, and building or customizing models. Tobias Macey, who also hosts the Data Engineering Podcast, runs technical interviews that often close by asking the guest about the biggest gaps in tooling, technology or training for AI systems.
The show began in June 2022 under the name The Machine Learning Podcast. Recent conversations cover agent architecture and governance, AI-assisted software development, evaluation, observability and compute infrastructure. Episode 80 came out on 7 October 2026.
| Detail | AI Engineering Podcast | Chain of Thought |
|---|---|---|
| Hosts | Tobias Macey | Conor Bronsdon |
| First episode | June 2022 (as The Machine Learning Podcast) | November 2024 |
| Format | One-guest technical interviews (every episode from October 2025 to October 2026) | Interviews with engineers, founders and operators shipping AI in production |
| Typical length | 59 minutes (median, October 2025 to October 2026) | 53 minutes (median, October 2025 to October 2026) |
| Transcripts | On episode pages | A full transcript on every episode page |
| Where to listen | aiengineeringpodcast.com | Episode archive and podcast apps |
Figures for the AI Engineering Podcast come from its own site and feed, checked 14 October 2026.
Chain of Thought is a podcast about AI infrastructure and the people building it. Conor Bronsdon interviews engineers, founders and operators shipping AI in production, and each conversation tends to stay on one company's decisions: what it built, what it cost and what it would change. Both are one-host interview shows; the AI Engineering Podcast describes itself as a guide to building and architecting AI systems, and many Chain of Thought guests are the product and engineering leaders deciding what to build.
Start with the AI Engineering Podcast if you want a technical interview about how a system is put together, from agent harnesses and MCP to RAG evaluation and GPU clouds.
Start with Chain of Thought if you want the decisions around the system: why a team picked a model, how it measured the result and what it cost.
You do not have to pick one. The pairings below are where the two shows meet.
Preeti Shukla discusses adding agents to multi-tenant SaaS while controlling latency, cost, privacy, access and autonomy. She also covers model routing, human review, evaluation and observability.
Then play
How Superhuman Built AI Into a 100ms Product | Loïc Houssier Loïc Houssier, Superhuman Loïc Houssier shows how Superhuman routed models and evaluated AI features inside a product built for sub-100ms speed while allowing auto-drafts but no auto-sends.Guillaume de Saint Marc of Cisco Outshift discusses open protocols for agent collaboration and tool calls, plus discovery, identity, observability and access control.
Then play
Architecting Reliable Agentic AI | Cisco’s Giovanna Carofiglio on the AGNTCY Collective Giovanna Carofiglio, Cisco Giovanna Carofiglio of Outshift by Cisco takes the interoperability question into AGNTCY, discussing agent discovery, identity, MCP, A2A and evaluation.Maxime Beauchemin discusses coordinating multiple coding agents through context as code and Agor, while code review, QA and asynchronous work become bottlenecks.
Then play
After Code Gen: What Graphite Is Building for the Post-AI Dev Stack | Greg Foster Greg Foster, Graphite Greg Foster of Graphite picks up that code review bottleneck and explains how teams handle the pull requests that AI coding tools now generate.Saturn Cloud co-founder and CTO Hugo Shi maps the GPU market, from hyperscalers and GPU clouds to bare-metal providers and aggregators, including orchestration, data gravity and how AMD's ecosystem is maturing as competition to NVIDIA.
Then play
Tuning GPU Performance with AI Agents | AMD’s Anush Elangovan on ROCm 10 Anush Elangovan, AMD Anush Elangovan of AMD follows the compute choice down to GPU workloads, showing how agents tune kernels with Hyperloom and ROCm 10.Ofer Mendelevitch discusses RAG evaluation when source data changes or lacks clean ground truth, including retrieval, access controls, synthetic questions and production traffic.
Then play
Explaining Eval Engineering | Galileo's Vikram Chatterji Vikram Chatterji, Galileo Vikram Chatterji of Galileo shows how human feedback can update an evaluator and how teams can run those checks at production scale rather than treating evals as a one-time test.TrueFoundry co-founder Nikunj Bajaj describes the agent harness around a model, including memory, tools, permissions, observability, guardrails and routing.
Then play
11 Cameras, Dozens of Mics: The AI That Reads the Room | Tormod Ree, Neat Tormod Ree, Neat Tormod Ree of Neat gives a hardware example of harness design, combining signals from edge models to make decisions inside a meeting room.Facts about the AI Engineering Podcast on this page were checked against these pages on 14 October 2026.