This answer is based on Morris Dworkin's report of the conference, so it ultimately relies on its accuracy (it is probably more reliable than memory, though).
Regarding the question "Except for your proposal for AES, what is your candidate?", the report states (page 15):
The panelists were asked which algorithm, other than their own, they
would choose for the AES. Rijmen liked RC6; the other four panelists
said Rijndael if it was extended to 18 or more rounds.
So this suggests that the representative for Rijndael (Vincent Rijmen) chose RC6.
During his presentation at the conference, Bruce Schneier is reported to have ended with the following suggestions (page 15 as well):
He [Schneier] recommended that NIST choose either Rijndael extended to 18
rounds, Serpent, or Twofish for the AES; he favored Twofish
for its efficiency across the board, its unique flexibility, and its
Since he suggested 18 rounds here, it would surprise me if he would argue for 32 rounds later the same day.
So the representative who answered "Rijndael with 32 rounds" was either Shai Halevi (MARS), Ron Rivest (RC6) or Ross Anderson (Serpent).
My personal guess would be Ross Anderson (this would make sense because Serpent has 32 rounds). In fact, on page 16 of the report there is some weak evidence pointing in this direction:
Anderson responded to a suggestion to transfer some of
Serpent’s rounds to Rijndael by reiterating his support for 32 round
Serpent with 256 bit keys.