@tdpham2 has some work on doing charge prediction without DFT. We should use it as a complementary approach to the DFT