I don't think GI/HDRI is any different from raytraced except the number of protons from different sources. You could build your own GI and HDRI just like you did with Baker. What you need is algorithms for light map to work with light sources in 2D platforms. It can be complex for 3D conversion to 2D light map on every objects. After you make this you will want to implement new Layers Displacement/Normal Map as other compression for renderer to cut down unnecessary polys and convert it into displacement/normal map for renderer only. Because single image is only 2D nothing more. It is very wasteful CPU resources for 3D renderer to do all the numbers to make a color pixel in the end.