A research lab building behavioral intelligence.

Every year, the advertising industry wastes $190 billion on campaigns that fail to connect with their intended audience. The core problem is simple: advertisers have no way to predict how their creative will perform before spending real money on media.

SimStudio changes this. We built an AI simulation engine that models how diverse human audiences react to video advertising — before a single dollar is spent. Think of it as the weather model for advertising: just as meteorologists simulate atmospheric conditions to predict storms, we simulate audience behavior to predict campaign performance.

Why we're different

Most AI tools for advertising rely on prompt engineering — asking a language model to “pretend to be a 35-year-old Chinese male luxury traveler.” This approach hits a fundamental accuracy ceiling around 60%, regardless of the model used. The reason is simple: language models are trained on internet text, not on actual human behavior data.

SimStudio takes a fundamentally different approach. We use steering vectors — mathematical representations learned directly from real-world behavioral datasets — to control AI personas in latent space. This is not prompt engineering. It is deterministic, verifiable, and calibrated against ground-truth data.

The Ads World Model

Our vision extends beyond simulating individual consumers. The advertising ecosystem is a complex network of users, content, brands, and social connections. SimStudio models this entire network using heterogeneous information graphs — the same mathematical framework our founding research pioneered with the HGT (Heterogeneous Graph Transformer) architecture.

We call this the Ads World Model: a complete simulation of how advertising propagates through, and is received by, human society. Stage I is our AI agent-based audience simulator. Stage II extends this to agent social networks — because the future of the internet isn't just humans consuming content, it's AI agents acting on behalf of humans.

Built as a research lab

Aveo Research Labs is a behavioral AI research lab with founders from UCLA, LinkedIn, Zscaler, and xAI alumni. Our published research spans 29,500+ citations across graph neural networks, behavioral foundation models, and causal inference. We are not an AI wrapper. We build fundamental infrastructure for the agentic economy.