Simile raises $2B Series B to build digital twins of human behavior for enterprise simulation

Simile AI, led by CEO Joon Sung Park, announced a $2 billion Series B funding round led by GreenOaks Capital Management and Index Ventures. The round includes prominent backers such as Fei-Fei Li and Andrej Karpathy.

The company states it runs tens of millions of simulations for Fortune 100 clients, including CVS, and claims its digital twins reproduce human behavior and attitudes with 85–99% accuracy compared to human focus groups.

Simile’s technology traces its origins to Park’s 2023 research on Generative Agents, which demonstrated AI characters capable of remembering, planning, socializing, and exhibiting emergent behaviors in a simulated environment called Smallville.

Park describes the company’s current focus as building foundation models of human behavior, moving beyond generative agents to create what he terms “digital twins” that aim to simulate entire populations.

Simile’s approach combines long-form interviews, observational and transaction data, randomized controlled trials, and post-training on causal mechanisms to model human decision-making at both population and individual levels.

Park argues that frontier models optimized for rationality may fail to capture the irrational behaviors central to realistic human simulations, advocating for changes to model weights rather than relying solely on prompting techniques.

The company’s stated ambition includes using simulations to test products and policies before deployment, identify counterintuitive pathways to desired outcomes, and model emergent behaviors across societies.

Park positions simulation as a new scaling law for AI, drawing parallels to Thomas Schelling’s work in agent-based modeling and suggesting that large-scale simulations may require data-center-scale compute infrastructure.