Meena Jagadeesan joined the CIS Department in July 2026 as an Assistant Professor. Her work lies at the intersection of machine learning and multi-agent systems. Rather than viewing an ML model as an isolated system, her research centers on modern machine learning ecosystems where an ML model interacts with humans, other models, and model-providers. Her work investigates how multi-agent interactions in ML ecosystems shape safety, performance, and society. 

A major focus of Jagadeesan’s work is on how multi-agent interactions lead to unanticipated failure modes as well as performance benefits that standard machine learning evaluations fail to capture. For example, her work characterizes how model-provider competition distorts scaling laws, how feedback loops with an ML model lead to reward hacking and supply-side effects, and when aggregating multiple copies of an ML model outperforms a single model. 

Using these characterizations, Jagadeesan’s research develops interventions that explicitly account for how multi-agent interactions shape ecosystem-level outcomes. This ranges from model-level interventions such as incentive-aware learning algorithms, to evaluation procedures that capture ecosystem-level metrics, to technical tools to assess policy-level interventions. Altogether, her goal is to develop the necessary foundations to steer machine learning ecosystems towards the desired outcomes.