There are a few diagrams in the original white paper (Anish Roshi et al. 2021; see Fig. 11) and the revised NGAT-130 proposal (Anish Roshi et al. 2023; see Fig. 3), depicting the basic arrangement. Here's Fig. 11 from the first paper, showing the telescope as viewed from the top (left image) and side (right image, shown with only a few dishes installed). The plate looks more like an oval than a circle, but I believe a circular plate would be more representative of the actual design.
Here's an artist's impression from a talk by some of the scientists behind the idea (see 54:30 for a broader set of designs and 55:05 for the then-final design). The blue dome isn't meant to be real, but rather to represent radio frequency interference mitigation techniques:
Some variations (like the one shown above) had the small dishes embedded in the plate, while others had them mounted firmly above it, but the basic principles are the same. Versions of this configuration are already in use on smaller scales; both AMiBA and Pluton have been using it for decades.
Both of these depict the original design of the NGAT: a 314-meter-wide plate with 1,112 9-meter dishes. However, this subsequently evolved into the NGAT-130, a 146-meter-wide plate with 102 13-meter dishes. While smaller, it would be more cost-effective and would still satisfy most of the scientific and technical requirements put forth in the white paper. It could also, potentially, be used as a precursor to a larger instrument like the original NGAT design.
One reason for the array-on-a-plate idea instead of a single dish is that it provides better sky coverage. For atmospheric studies, the NGAT would need to cover zenith angles as large as 45$^{\circ}$, which is very hard to do with a single dish, even with a deformable secondary reflector. There were also issues of limited transmitting power for radar studies. The coverage issue could be circumvented by a tight array of small dishes, but changing orientations of dishes relative to one another in such an arrangement would make it difficult to predict the radiation pattern due to interactions between antennas. It turns out that small dishes mounted on a steerable plate and oriented in the same direction allow for larger zenith angles (up to 80$^{\circ}$ with the NGAT-130!) without the pattern issues.
The NGAT site could be in a variety of places and the design doesn't necessitate that it be built at Arecibo; however, natural RFI shielding from the karst hills around the site and the fact that it is located in a special RFI coordination zone mean that it has a leg up on other options.
