Why are we doing this?

The data that surrounds homelessness is fragmented, disconnected, dated, or in most cases does not exist. Homelessness affects everyone where it occurs and without a clear, up-to-date picture on the situation there is no verifiable way to track what resources, interventions, and treatments are having an impact. We want to change that by democratizing outreach and enabling everyone to help us build the clear picture required to effectively utilize and implement resources and initiatives. 

How can I submit data?

There are a few easy ways to submit data to Nomadik:

  1. Right click on the map, or use the point dropper, on a spot on the map you have information on, type it in and send it our way.
  2. Email info@nomadik.ai images, videos, or any other information you may have with a location.
  3. Subscribe to get on the waitlist for our soon to be released Android and iOS app!

How are we using AI/ML to inform solutions?

Contrary to popular belief, there is not one centralized machine learning network, or AI, performing the analysis seen on our website, but a multitude of small, purpose built models. This strategy enables us to use isolated and domain agnostic data to train individual models to complete small tasks, simultaneously breaking down the complexity of the bigger picture, and ensuring robustness and the isolation of sensitive data. 

These specialized models allow us to build an accurate picture of features such as camp boundaries, debris levels, hazard levels, and more. Given that submitted data is location tagged point-in-time information we are able to feed the state of the region over time to further models that forecast specifics such as impact of outreach and available resources or hazard impacts to local areas. 

How do we account for the privacy of individuals?

All submitted data is securely encrypted and stored for processing by our models. Individual privacy is maintained across all processing techniques we apply by encoding all media we receive through a process called vectorization, the results of which are only useful to the models themselves.