Research

Do you know what the arrival time p-value of an Uber trip?

We all rely on knowing how long it takes to get somewhere, but with traffic and other factors, those "estimated" arrival times aren't always reliable. What if you could know the precise odds of being delayed by X minutes? 

A p-value, used in statistics, helps answer exactly that! It measures the probability of something happening – in this case, a longer-than-expected trip. To figure this out, we need to understand how travel times are distributed.

My work establishes an interesting phenomenon of travel time in real-world networks, which enable us to quantify exactly the likelihood of delay by X minutes in real-world data. Published at the Annals of Applied Statistics [more].

Can we sample/simulate rides independently of the road-graph? 

Transportation systems are indeed physical systems, with ride-share providers (like Lyft, Uber, Careem) and autonomous vehicles (Waymo, Zoox, Nuro) relying on road networks (or metric graphs) for navigation. Physics-based traffic simulators offer valuable insights, but if our goal is to analyze the statistical properties of rides within these systems, the dependence on specific road graphs can be limiting.

This write-up introduces a sampling method designed to preserve the statistical properties of ride populations while being independent of the road network. [more]

Predictive modeling for pandemic prevention: Can big data save us? 

In the ongoing data revolution of biodiversity research, global databases mapping species interactions are growing rapidly, but many interactions remain undocumented. Predicting these missing links in ecological networks is key to understanding biodiversity and tracking pathogens that infect multiple hosts.

My research, focuses on utilizing host phylogeny to improve these predictions. By incorporating evolutionary relationships between species into network-based models, we can reduce uncertainty about unknown interactions and enhance our understanding of ecological communities. This novel approach has proven valuable in collaborations with ecology research groups, leading to publications in the Annals of Applied Statistics, Journal of Animal Ecology and the American Journal of Biological Anthropology. [AOAS, JAE, AJBA]