Best AI Engineering Podcasts for Builders in 2027

Developer desk at night with earbuds and a glowing terminal in cool blue light

The 11 best AI engineering podcasts for 2027, from Latent Space and Practical AI to Vanishing Gradients, with cadence, episode length and audience numbers.

AI / AI Education / AI podcasts

The best AI engineering podcasts for 2027 are Latent Space for the AI engineer community, Practical AI and TWIML for applied case studies, and Agentic Conversations and Vanishing Gradients for agents and evals in production. The Pragmatic Engineer Podcast and Dev Interrupted cover what AI does to engineering teams. We picked 11 builder shows that published new episodes in the 90 days before 19 November 2026, grouped them by the problem they help with, and list each show's cadence, typical episode length and published sponsorship contact.

Key takeaways

  1. Latent Space is the default first subscription for AI engineers; add one agent-focused show and one team-focused show.
  2. Episode length ranges from 30 minutes (Stack Overflow) to more than two hours (Vanishing Gradients), so match the show to your listening window.
  3. Two shows here, Dev Interrupted and Agentic Conversations, are backed by a vendor or foundation; the guests are the reason to listen.

Which AI engineering podcasts are worth a builder's time?

Ranked by how much a practitioner or technology leader gets out of each one. "Checked" names the source behind each number and when we looked.

Core shows for AI engineers

  1. 545K YouTube subscribers · 4.5 on Apple (108 US ratings) · about 3 episodes a week, often 80 min or more

    For: AI engineers and platform teams who build agents, retrieval systems and model-backed features.

    Latent Space covers the full AI engineering stack, from inference and coding agents to forward-deployed engineering, in about three episodes a week. Its separate feeds for engineering, science and short Lightning episodes let a listener subscribe only to the part they need. Recent guests included Black Forest Labs and Resend, whose episode looked at designing developer products for agents as users.

    Checked: Apple Podcasts US page (4.5, 108 ratings), RSS feed (37 episodes in 90 days), YouTube channel page and latent.space/about, 19 Nov 2026

  2. 4.3 on Apple (202 US ratings) · weekly, about 45 min · 380 episodes

    For: Developers and data scientists who want applied AI explained without hype or heavy math.

    Practical AI has published 380 weekly episodes on implementation, and its recent run covered federated learning for agents and agentic vulnerability discovery. The hosts, Daniel Whitenack and Chris Benson, keep episodes to about 45 minutes. It is the easiest show here to recommend to an engineer moving into AI work for the first time.

    Checked: Apple Podcasts US page (4.3, 202 ratings), RSS feed and practicalai.show, 19 Nov 2026; email is the owner address in the RSS feed

  3. 31.1K YouTube subscribers · 4.7 on Apple (427 US ratings) · weekly, about 55 min · 785 episodes

    For: ML engineers who want case studies of how named companies build and run AI systems.

    Sam Charrington's weekly interviews now focus on enterprise builds: recent episodes covered Airbnb's knowledge layer for agents, Canva's AI design work and the economics of buying tokens versus building an AI factory. Each episode is about an hour. The 785-episode back catalog is searchable by topic on twimlai.com, which makes it useful as a reference library.

    Checked: Apple Podcasts US page (4.7, 427 ratings), RSS feed, YouTube channel page and twimlai.com/contact (sponsorship topic), 19 Nov 2026

Agents, evals and production

  1. 4.6 on Apple (25 US ratings) · about 2 episodes a week, about 35 min

    For: Engineers running agents in production who want short, practical conversations about what breaks.

    Formerly the MLOps.community podcast and now titled Agentic Conversations, Demetrios's show is backed by the Agentic AI Foundation, home of MCP, AGENTS.md and Goose. Episodes run about 35 minutes, twice a week. Recent topics included how Stripe cut 500,000 internal documents down to 5,000 for agents and moving from slop code to software factories.

    Checked: Apple Podcasts US page (4.6, 25 ratings), RSS feed and home.mlops.community, 19 Nov 2026; email is the owner address in the RSS feed

  2. 6K YouTube subscribers · 5.0 on Apple (14 US ratings) · 1 to 2 episodes a month, often 60 to 140 min

    For: Senior engineers and data scientists who want long, careful conversations about agents and evals.

    Hugo Bowne-Anderson's long-format interviews cover agents, evals, multimodal systems and data infrastructure, with past guests including Jeremy Howard, Hamel Husain, Shreya Shankar and Wes McKinney. Recent episodes included building a coding agent with the Amp Code team. It publishes once or twice a month, and episodes often pass two hours.

    Checked: Apple Podcasts US page (5.0, 14 ratings), RSS feed and YouTube channel page, 19 Nov 2026; no sponsorship contact published

  3. 4.4 on Apple (7 US ratings) · 1 to 3 episodes a month, about 60 min

    For: Engineers who want architecture-level episodes on building maintainable AI systems.

    Tobias Macey's show positions itself as a guidebook to building scalable and maintainable AI systems. Recent episodes covered evaluating RAG systems when ground truth keeps changing and harness engineering for governed agents. It publishes one to three hour-long episodes a month, so it fits as a slow, deep second subscription.

    Checked: Apple Podcasts US page (4.4, 7 ratings) and RSS feed (email in show notes), 19 Nov 2026

  4. 4.5 on Apple (31 US ratings) · weekly, about 50 min

    For: Data and AI engineers who want weekly conversations on data infrastructure for agents.

    Ben Lorica's show, produced by Gradient Flow, pairs interviews with roundups of data and AI news. Recent episodes asked why agents need a new data stack and who proves AI-written code works. Episodes run about 50 minutes and lean toward data platform and governance questions other builder shows skip.

    Checked: Apple Podcasts US page (4.5, 31 ratings), RSS feed and thedataexchange.media, 19 Nov 2026

Engineering teams and leadership

  1. 545K YouTube subscribers · 4.9 on Apple (102 US ratings) · weekly, often 90 min

    For: Engineers and engineering managers who want to understand how AI changes teams, code review and careers.

    Gergely Orosz's interviews come from inside engineering organizations, and the recent run shows how central AI has become: Satya Nadella on engineering after code becomes cheap, Harjot Gill on software factories and code reviews, and the team behind Uber's AI infrastructure. Episodes often run 90 minutes. The 545K-subscriber YouTube channel exists, but the format suits audio.

    Checked: Apple Podcasts US page (4.9, 102 ratings), RSS show notes (sponsor email), YouTube channel page and podcast sponsors page, 19 Nov 2026

  2. 4.4 on Apple (622 US ratings) · about 2 episodes a week

    For: Engineers who want a broad weekly feed of infrastructure and software interviews with steady AI coverage.

    Founded in 2015 by Jeff Meyerson, Software Engineering Daily publishes about two episodes a week for an audience it describes as developers, founders and technology leaders. AI tooling, agents and inference now appear in much of the schedule. It is wider than the AI-only shows here, which helps engineers keep AI in context with the rest of the stack.

    Checked: Apple Podcasts US page (4.4, 622 ratings), Apple episode dates and the show's sponsor page, 19 Nov 2026

  3. 4.6K YouTube subscribers · 4.8 on Apple (141 US ratings) · Tuesday interview plus Friday news roundup, about 45 min

    For: Engineering managers and directors deciding how AI coding tools change process, quality and governance.

    LinearB's show runs a Tuesday interview and a Friday news roundup focused on agentic development and AI-native engineering. Recent guests discussed balancing AI with compliance at Capital One, accountability for AI-generated code at Freshworks and AI bills of materials. It is vendor-produced, and its value is in the guests rather than the product.

    Checked: Apple Podcasts US page (4.8, 141 ratings), RSS feed, YouTube channel page and linearb.io, 19 Nov 2026

  4. 4.3 on Apple (64 US ratings) · about 2 episodes a week, about 30 min

    For: Developers who want short, frequent episodes on how AI meets everyday software work.

    Hosted by Ryan Donovan, the Stack Overflow Podcast publishes about two 30-minute episodes a week. Recent episodes covered why agents need OAuth2, how agentic search assembles answers and why customer support is a hard AI problem. Short episodes make it the easiest way to keep a developer audience current between longer shows.

    Checked: Apple Podcasts US page (4.3, 64 ratings), RSS feed and Stack Overflow advertising page, 19 Nov 2026

What makes a good AI engineering podcast?

A good AI engineering podcast talks to the people who built the system, at a level of detail that changes how a listener builds the next one. The shows on this list share that trait: guests explain architecture choices, eval methods and failure modes rather than product roadmaps. Recent episodes on RAG evaluation, agent harnesses and code review for AI-written code are typical.

Recency matters more in this niche than in any other. Tooling for agents and inference changes month to month, so we excluded shows without new episodes in the last 90 days, which removed Gradient Dissent and AI + a16z despite strong back catalogs. Older episodes from those shows can still be worth playing for fundamentals.

We also weighted range. Latent Space and TWIML cover the whole stack, Vanishing Gradients and the AI Engineering Podcast go deep on a few topics, and the Pragmatic Engineer Podcast and Dev Interrupted look at the organizational side. A builder's stack should mix at least two of those angles.

How should an AI engineer build a weekly listening stack?

Plan for about two hours a week. Latent Space alone can fill three if you play everything, so pick its episodes by guest and use the separate engineering feed if science episodes are not relevant. Add one weekly applied show, such as Practical AI or TWIML, for case studies from outside the frontier labs.

Next, add one show that tracks your current problem. Teams putting agents into production should take Agentic Conversations or Vanishing Gradients; data platform teams will get more from The Data Exchange. Rotate this slot when the problem changes, and unsubscribe from shows you have not opened in a month.

Engineers moving into lead roles should add The Pragmatic Engineer Podcast or Dev Interrupted. Both treat AI as a team and process question, which is the question a lead gets asked. For written companions to these shows, see our guide to building an AI-native engineering team and our coverage of agentic software development.

How did we check these AI engineering podcasts?

Every figure on this page comes from a source we opened on 19 November 2026. Ratings are the US storefront counts on each show's Apple Podcasts page, YouTube numbers are channel subscriber totals, and cadence and episode length come from counting episodes in each show's RSS feed over the previous 90 days. Small rating counts, such as the AI Engineering Podcast's seven, say more about how the audience listens than about quality.

Where a show publishes a sponsor address on its site or in show notes, the listing carries it; otherwise the listing uses the RSS owner address and the checked note says so. Software Engineering Daily and the Pragmatic Engineer Podcast publish dedicated sponsor addresses; Vanishing Gradients publishes none, so its field is empty. We did not estimate downloads for any show, because none of these builder podcasts publishes a verified download figure.

Listed here, or should be?

If an entry is out of date or something belongs on this list, email [email protected]. Listings are editorial and free. A featured, labeled Sponsored slot is available on request.

Related questions

What is the best podcast for AI engineers?

Latent Space is the strongest single choice. It covers agents, inference, coding tools and evals with the people building them, publishes about three episodes a week and splits its catalog into separate feeds for engineering, science and short episodes.

Which AI podcast covers agents in production?

Agentic Conversations, formerly the MLOps.community podcast, and Vanishing Gradients focus most on agents in production. Agentic Conversations runs short twice-weekly episodes; Vanishing Gradients publishes long, detailed conversations once or twice a month.

Is there a short AI podcast for developers?

The Stack Overflow Podcast runs about 30 minutes per episode, twice a week, and regularly covers AI topics such as agent authentication and agentic search. Interconnects also publishes short audio essays, though it leans toward research.

Explore More

Ready to Find the Right AI Tools?

Browse our data-driven rankings to find the best AI tools for your team.