LLM Search, UI/UX challenges, Context Engineering and the 80/20 of Eval

Episode Link

https://share.snipd.com/episode/3e2e6726-4d63-4a9a-8211-c38caa92e238

Episode publish date

September 5, 2025 2:01 PM (UTC)

Last edit date

Sep 14, 2025 9:10 AM

Last snip date

September 13, 2025 12:55 PM (GMT+1)

Last sync date

September 14, 2025 10:07 AM (GMT+1)

Show

MLOps.community

Show notes link

https://podcasters.spotify.com/pod/show/mlops/episodes/LLM-Search--UIUX-challenges--Context-Engineering-and-the-8020-of-Eval-e37qo12

Snips

Warning

⚠️ Any content within the episode information, snip blocks might be updated or overwritten by Snipd in a future sync. Add your edits or additional notes outside these blocks to keep them safe.

‣

Episode show notes

Your snips

‣

[00:27] Real Agents: Tukan And OLX

‣

[02:20] Context Beats Model Choice

‣

[06:26] Pipeline Live Context Into Prompts

‣

[11:23] Search vs Context: Different Roles

‣

[15:11] Use LLMs Before And After Search

‣

[18:27] When Chatbot Demos Flopped

‣

[22:58] Handhold Users For New UIs

‣

[26:56] Generative UI For Visual Decisions

‣

[29:03] Let Agents Watch Interactions (With Consent)

‣

[32:44] Send Few, Timely, Personalized Prompts

‣

[36:16] Evals Are The Real Moat

‣

[38:56] Run Evals Early With Synthetic Data

‣

[40:49] Use Deterministic Funnel Metrics First

‣

[43:47] Calibrate LLM Judges With Labeling Parties

‣