New Technology / Ai Development

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

Criteo utilizes advanced AI techniques to enhance personalized advertising, which is crucial for supporting small businesses and niche products online. The company operates under strict time constraints, processing user data and recommending products in milliseconds through a modular system that employs multiple foundation models. Criteo employs anonymous IDs to track user interests without collecting personal information, enhancing the relevance of advertisements. This approach aims to improve user experience while supporting the open internet through advertising revenue.
← back to ALL
cognitive_revolution_how_ai_changes_everything • 2026-05-09T13:00:14Z
Source material
Summary
Criteo utilizes advanced AI techniques to enhance personalized advertising, which is crucial for supporting small businesses and niche products online. The company operates under strict time constraints, processing user data and recommending products in milliseconds through a modular system that employs multiple foundation models. Criteo employs anonymous IDs to track user interests without collecting personal information, enhancing the relevance of advertisements. This approach aims to improve user experience while supporting the open internet through advertising revenue. Criteo utilizes advanced AI and machine learning models to enhance targeted advertising by analyzing user behavior and preferences. The partnership with OpenAI aims to improve user profiling and ad relevance while addressing privacy concerns. Criteo is integrating large language models with its commerce data systems to enhance product recommendations and user experiences. The company faces challenges in aligning dynamic commerce data with the static training data of these models.
Perspectives
LLM output invalid; stored Stage4 blocks + metrics only.
Metrics
50 people
size of the AI Lab team
A larger team can enhance research output and innovation
they're still comfortable publishing the AI Lab's full 50-person roster to their website.
between 200 to the 1000 features
range of features in modern deep learning models
This shift indicates a significant enhancement in the model's ability to process and analyze data
now essentially we have something like between 200 to the 1000 features
300%
performance during peak times
This indicates the extreme demands placed on the system during high-traffic periods
imagine taking your car and running it for one week at 300 miles an hour
50 faces on the website people
size of the AI lab team
A larger team may indicate a stronger capacity for innovation and development
I think it's like I think I saw like 50 faces on the website
over 20 years
duration of Criteo's commitment to building a great workplace
Long tenure indicates stability and employee satisfaction
we've been doing a great job for over 20 years now
17,000 units
of advertisers using Criteo's platform
A larger advertiser base indicates Criteo's market reach and effectiveness
I think the number I saw was credo has 17,000 advertisers
Key entities
Companies
Anthropic • Criteo • OpenAI • Stripe • Waymark
Countries / Locations
ST
Themes
#ai_development • #big_tech • #ad_performance • #ad_targeting • #ad_tech • #ad_technology • #adtech • #advertising
Key developments
Phase 1
Criteo utilizes advanced AI techniques to enhance personalized advertising, which is crucial for supporting small businesses and niche products online. The company operates under strict time constraints, processing user data and recommending products in milliseconds through a modular system that employs multiple foundation models.
  • Criteo leverages advanced AI techniques to enhance personalized advertising, which is vital for supporting small businesses and niche products online
  • The company operates under strict time constraints, processing user data and recommending products in milliseconds through a modular system that utilizes multiple foundation models
  • Criteos partnership with OpenAI seeks to merge real-time product inventory data with ChatGPTs capabilities, potentially revolutionizing product discovery
  • The company emphasizes its commitment to privacy and compliance, asserting that European regulations are manageable and highlighting the strength of the European AI talent pool
  • The discussion explores the future of advertising, suggesting that AI agents may increasingly take on roles in product discovery, altering the traditional value exchange as human time becomes more precious
Phase 2
Criteo employs anonymous IDs to track user interests without collecting personal information, enhancing the relevance of advertisements. This approach aims to improve user experience while supporting the open internet through advertising revenue.
  • Criteo prioritizes transparency in data collection, using anonymous IDs to track user interests without gathering personal information
  • Personalized advertising is argued to improve user experience by filtering out irrelevant content, leading to a more engaging environment for consumers
  • Advertising is essential for sustaining the open internet, as it generates revenue for content and service providers, enabling free access to their offerings
  • Criteo emphasizes user choice through its involvement in the AdChoices program, which informs users about ad visibility and provides opt-out options
  • The evolution of data collection practices is influenced by changes from companies like Apple, which have reshaped cookie management and user data handling in advertising
Phase 3
Criteo utilizes advanced AI and machine learning models to enhance targeted advertising by analyzing user behavior and preferences. The partnership with OpenAI aims to improve user profiling and ad relevance while addressing privacy concerns.
  • Criteo employs cookies to anonymously track user interests, enabling targeted advertising based on users previous online behavior
  • Machine learning models are used to assess ad placement effectiveness by predicting user engagement, balancing accuracy with the need for model transparency
  • The partnership between Criteo and OpenAI focuses on enhancing user profiling and ad relevance through advanced conversational AI, while addressing ongoing privacy concerns
  • A key challenge in advertising technology is creating sophisticated models that improve targeting without compromising transparency and user understanding
Phase 4
Criteo is integrating large language models with its commerce data systems to enhance product recommendations and user experiences. The company faces challenges in aligning dynamic commerce data with the static training data of these models.
  • Criteo is integrating large language models (LLMs) with its commerce data systems to improve product recommendations and user experiences
  • A significant challenge is aligning the frequently updated commerce data with the static training data of LLMs to ensure users receive accurate product information
  • Criteos network of 17,000 retailers enables real-time data ingestion, essential for maintaining current product availability and pricing, particularly during high-demand events like Black Friday
  • The proposed hybrid architecture combines LLM conversational capabilities with precise commerce data, aiming to enhance the shopping experience
  • This integration seeks to mimic the personalized assistance of knowledgeable sales staff in physical stores, offering tailored recommendations and insights
Phase 5
Criteo is leveraging advanced AI and machine learning to enhance targeted advertising through real-time data processing and user profiling. The integration of large language models aims to improve product recommendations while addressing privacy concerns.
  • Modern billing systems for AI products are complex, necessitating comprehensive revenue workflows that can manage intricate pricing models and automate finance tasks
  • Claude, an AI tool by Anthropic, enhances productivity by assisting users in organizing personal data and performing tasks like tax preparation and drafting investment memos
  • The architecture of ad tech models requires rapid processing speeds to effectively deliver ads, which involves handling large volumes of user profiles and advertiser data
  • Effective input management for ad tech models is crucial, with tokenization being essential to condense user profiles into a manageable format for real-time ad matching
Phase 6
Criteo employs advanced AI and machine learning to optimize ad placements by analyzing user behavior and preferences. The integration of deep learning allows for a more nuanced understanding of user data, enhancing predictive accuracy.
  • Criteos ad tech assesses the value of advertising opportunities by analyzing user behavior, including clicks and purchase history, to inform bidding strategies
  • The system employs approximately 150 features, such as product interest and website context, to predict user engagement and optimize ad placements in real-time
  • Criteo has transitioned from simpler models like logistic regression to deep learning, enhancing feature extraction and predictive accuracy
  • This shift to deep learning enables a more nuanced understanding of user data, allowing models to autonomously compute relevant features rather than relying on manually crafted ones
  • Criteos advanced models now manage a significantly larger array of features, improving their effectiveness in matching ads to users

Related coverage

Adjacent technology themes

Commercialization and strategic context