Decentralized wastewater systems (e.g., onsite wastewater treatment systems, septic tanks) can contribute significantly to nutrient fluxes in coastal and estuarial waters.  Despite the potential impact of these systems, their contributions to nutrient flux are not well quantified. One of the challenges in quantifying these impacts is that there is often poor geospatial data available on where these systems are located and how many of them are in use. Where records are publicly available, they are rarely comprehensive, often stored in legacy formats, and difficult to access in bulk. This lack of accessibility has resulted in a limited ability by environmental managers, researchers, and other interest holders to  quantify the impacts of decentralized wastewater systems as they relate to a range of topics including nutrient fluxes. There are also gaps in data availability for locations served by centralized sewer systems, which can make accurately identifying what type of wastewater service a home has more difficult. Methods in machine learning present a unique pathway for overcoming these data gaps. We present a machine learning model developed for identifying wastewater infrastructure at the land parcel scale, with a focus on model performance in the Gulf region, and the resulting inventory of wastewater system types in the region. The model consistently demonstrates accuracy greater than 80% in states in the Gulf Region. We also present an object detection model for identifying locations of utility hole covers with Google Street View imagery. The model achieved F-1 scores between 0.92-0.96 across 5 spatially distinct held-out test sets in Florida. These machine learning methods serve as elements in an emerging suite of tools for characterizing the extent to which different types of wastewater systems are distributed across the Gulf region and addressing key questions relating to wastewater infrastructure access and environmental impacts.