About Aaru
Aaru builds simulations of human behavior. Each simulation contains a population of AI agents, each representing a person who could plausibly exist in the real world and capable of making decisions within a modeled environment. Companies and institutions use these simulations to test consequential choices before committing—from product launches and pricing decisions to strategic communications and policy changes.
Building a useful simulation requires more than generating plausible text. Populations must represent real people and groups; predictions must be calibrated; simulations must remain coherent as conditions change; and the product must make the resulting evidence legible enough to support real decisions.
We are a small, in-person team in New York. We work with urgency, high ownership, and intellectual honesty. We expect people to surface inconvenient evidence, change their minds quickly, and carry important work all the way to a result.
About Simulation Engineering
Simulation Engineering owns the path from a promising research result to a production system that customers can trust. Researchers may prove a new method for constructing a population, modeling a world, estimating an outcome, or evaluating fidelity. Simulation Engineering turns that method into robust, reusable, observable, and efficient software.
The team's quality bar is broader than conventional service reliability. A production simulation must be behaviorally faithful, calibrated, reproducible, measurable, fast enough to use, economical enough to scale, and reliable under real customer workloads. The team owns the abstractions, contracts, evaluation gates, workflows, and operating systems that make those properties possible.
Simulation Engineering is not a research-support queue. It is the engineering owner of the simulation system in production. When a simulation method breaks, regresses, becomes too expensive, or produces conclusions that cannot be explained, this team is accountable for finding the cause and restoring trust.
How we work
We treat model and simulation behavior as production behavior. An unexplained regression is not an acceptable mystery; it is an incident to be measured, localized, and converted into a lasting test. A method is not ready because it produced an impressive example. It is ready when the evidence supports the intended claim, its limits are known, and the production system can reproduce and observe it.
We design evaluations before trusting a result. Strong baselines, ablations, temporal separation, subgroup analysis, and prospective outcomes matter. We optimize cost and latency only in relation to quality, and we refuse optimizations that make a system faster by making its uncertainty or errors harder to see.
Management is hands-on and high-context. The goal is to create clear interfaces, strong technical leaders, and an operating system that lets the team move faster with less hidden risk—not to centralize every decision in the manager.