Liftoff Expands Cortex AI Engine With New Predictive Modeling Features

Liftoff has introduced a new set of innovations for its Cortex neural network prediction engine, announced in a press release. Cortex powers all bidding and performance decisions across Liftoff's mobile advertising platform and now processes more than two billion predictions per second.

The latest update adds several modeling features. Unattributed sample learning expands Cortex's training data by including conversions not directly linked to its own campaigns. Multicast modeling enables user level return on ad spend optimization, shifting from aggregate predictions to individual user valuation. Sequential modeling incorporates full behavioral sequences such as timing, location, and app context to build a more detailed understanding of user intent.

These features are available to Liftoff Accelerate customers and have contributed to a 41 percent increase in revenue from the company's Core Advertising platform. Cortex has also improved experiment speed fourfold and reduced campaign optimization time from two weeks to less than a day.

Liftoff reports continued market adoption of Cortex, with a 21 percent year over year increase in demand side customers and more than 167,000 software development kit integrations reaching 1.4 billion daily active users as of March 2026. The company said it is now developing agentic workflows that will automate campaign testing and optimization using Cortex's real time prediction capabilities.