Ankur Goyal: Production AI Engineering starts with Evals
11 October 2024; 117 minutes
Foundational evaluation and product iteration discussion
Takeaways & episode link →SECTION 07 OF 12 / Apply & learn
Video and podcast roundups
Choose the conversations that address your next engineering decision.
- Pages
- 13
- Sources cited
- 7
- Last updated
- 26 October 2026
THE LISTENING LIST
Compare all five episodes →Choose the question first.
Philip Kiely and Ali Taha: The Inference Engineering Masterclass
3 August 2026; 101 minutes
Serving configuration, caching, precision and workload trade-offs
Takeaways & episode link →Diogo Almeida: Jev — System One models for Prod, not God
21 September 2026; 141 minutes
Task decomposition, calibration and service primitives
Takeaways & episode link →Erik Bernhardsson: Why AI Compute Comes in Rectangles
22 October 2026; 86 minutes
Variable demand versus fixed compute commitments
Takeaways & episode link →Evan Conrad: The Programmable Data Center
20 October 2026; 37 minutes
Delivery constraints behind apparently interchangeable GPU hours
Takeaways & episode link →Publisher pages and selected transcripts underpin the roundup. These conversations are not independent tests of your production system.
THE COMPLETE SECTION
13 PAGESRead the full research.
01Read →02Read →03Read →04Read →05Read →06Read →07Read →08Read →09Read →
What should a busy technical buyer listen to first?
Start with evaluation if you cannot define success; start with serving economics if a credible candidate already exists.
Which episode answers which decision?
The dates/durations come from the linked publisher records below.
1. Production AI Engineering starts with Evals — Ankur Goyal
Listen/watch: Latent Space episode page with audio, transcript and video link.
2. The Inference Engineering Masterclass — Philip Kiely and Ali Taha
Listen/watch: Latent Space episode page with transcript and YouTube link.
3. Jev: System One models for Prod, not God — Diogo Almeida
Listen: Latent Space episode and transcript.
4. Why AI Compute Comes in Rectangles — Erik Bernhardsson
Listen: Latent Space episode and transcript.
5. The Programmable Data Center — Evan Conrad
Listen/watch: Latent Space episode page with audio, transcript and video link.
How do we turn listening into a useful team session?
Assign one decision and one output to each participant, rather than asking everyone to consume every episode.
What should we avoid carrying forward?
Do not copy historical prices, fundraising totals or model rankings from a recording into a current business case.
10 / REFERENCE
Frequently asked questions
Short answers to the questions buyers ask.
Read →11 / REFERENCEWhat remains unproven
Evidence gaps and limits that still need resolving.
Read →12 / REFERENCEEditorial self-assessment
The research’s own quality scores and limitations.
Read →13 / REFERENCESources
Trace the section’s claims to the original sources.
Read →RESEARCH SELF-ASSESSMENTUseful evidence.
Useful evidence.
Visible limits.
9/10buyer fit
8/10depth
8/10search terms
7/10sources
Specific, dated selections with concrete buyer exercises and honest access limits. Strong relevance, but concentrated publisher coverage and no full audiovisual review limit the source score.