Current Work

Can a single cell learn the temporal structure of its environment? This work uses controlled cue–stress relationships in Capsaspora owczarzaki to distinguish reactive responses from preparation based on predictive information.

The broader aim is to connect measurable survival strategies with formal models of inference: what information is available, how it is integrated through time, and when acting on that information becomes advantageous.

Most microbial diversity can be sequenced but not grown under controlled conditions. This project treats cultivation as a history-dependent learning problem: environmental cues, stresses, recovery periods, and partner communities form paths rather than isolated settings.

The working platform combines inexpensive programmable chambers, live phenotypic measurements, and models that choose informative next experiments. Its success criterion is deliberately practical—stable cultures in fewer physical experiments than matched expert, random, or static-optimization approaches.

Using single-cell measurements, lineage barcoding, and population-level fitness assays in Saccharomyces cerevisiae, this work asks how the character and familiarity of a stressor change the adaptive possibilities available to a population.

The results connect environmental history, expression variability, and selection: familiar stresses can expose differentiated lineage-level strategies, while unfamiliar stresses may constrain populations to a smaller set of generic responses.