# Milliseconds to Match: Criteo's AdTech AI & the Future of Commerce w/ Diarmuid Gill & Liva Ralaivola

Page: https://stenobird.com/podcast/the-cognitive-revolution/milliseconds-to-match-criteo-s-adtech-ai-the-future-of-commerce-w-diarmuid-gill-liva-ralaivola
Text version: https://stenobird.com/podcast/the-cognitive-revolution/milliseconds-to-match-criteo-s-adtech-ai-the-future-of-commerce-w-diarmuid-gill-liva-ralaivola.md
Podcast: ["The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis](https://stenobird.com/podcast/the-cognitive-revolution)
Published: 2026-05-09T13:00:00+00:00
Episode link: https://www.cognitiverevolution.ai/milliseconds-to-match-criteo-s-adtech-ai-the-future-of-commerce-w-diarmuid-gill-liva-ralaivola/
Audio file: https://pdst.fm/e/mgln.ai/e/1113/pscrb.fm/rss/p/traffic.megaphone.fm/RINTP1840766402.mp3?updated=1778246281
Processing state: processed
JSON: https://stenobird.com/v1/public/podcasts/the-cognitive-revolution/episodes/milliseconds-to-match-criteo-s-adtech-ai-the-future-of-commerce-w-diarmuid-gill-liva-ralaivola
Duration seconds: 5235

## Resource

Explore the engineering required to run deep learning models and real-time bidding in milliseconds. Learn how Criteo uses embeddings and foundation models to power personalized commerce on the open internet.

## Highlights
- Main idea: Modern ad tech relies on high-speed recommendation systems that must process billions of profiles in milliseconds
- Technical takeaway: Effective real-time bidding requires a balance of pre-computed embeddings and lightweight runtime updates
- Practical takeaway: The partnership between Criteo and OpenAI aims to combine LLM world knowledge with real-time product inventory
- Failure mode: Personalized advertising fails if it becomes 'creepy' or lacks transparency, making user trust and privacy compliance essential
- Future trend: AI agents and conversational interfaces will shift product discovery from manual search to automated research

## Topics

AdTech, Deep Learning, Real-time Bidding, Embeddings, OpenAI, Privacy, E-commerce, Machine Learning Engineering

## Chapters
- 1:00 — The Value of Personalized Advertising: How recommendation systems support the open internet and small businesses.
- 8:00 — Privacy and User Control: The mechanics of user profiles and the ease of opting out of personalization.
- 14:30 — The Role of Conversational Agents: How LLMs and agents are changing the landscape of data privacy and interaction.
- 27:40 — Engineering Semantic Embeddings: The technical challenge of encoding products and users into meaningful vectors.
- 34:30 — Real-time Inference and Architecture: Managing the trade-off between model complexity and millisecond-speed execution.
- 41:10 — Vector Similarity and Product Discovery: Using embeddings to recommend similar products based on vector proximity.
- 1:01:20 — Global Privacy and EU Regulation: Navigating GDPR and the impact of European AI regulations on global tech stacks.

## Actions

- request_transcript: `POST https://stenobird.com/v1/public/podcasts/the-cognitive-revolution/episodes/milliseconds-to-match-criteo-s-adtech-ai-the-future-of-commerce-w-diarmuid-gill-liva-ralaivola/transcription-requests` — Idempotently request low-priority transcript generation for this episode.
- read_markdown: `GET https://stenobird.com/podcast/the-cognitive-revolution/milliseconds-to-match-criteo-s-adtech-ai-the-future-of-commerce-w-diarmuid-gill-liva-ralaivola.md` — Read the agent-friendly Markdown representation of this episode resource.

A page view does not enqueue transcription. Agents should invoke `request_transcript` explicitly when they need this episode processed.

## Transcript

Full transcripts are not published on public pages unless there is a clear rights basis.