... more than once per DNA sequence.

Maybe:

>perl -wMstrict -le
"my $seq  = 'atcgcgtacatcgatac';
 my $rule = '01234567890123456';
 my @uc_offsets = (0, 8, 15);
 ;;
 print qq{$rule};
 print qq{$seq};
 for my $offset (@uc_offsets) {
   substr($seq, $offset, 1) = uc substr $seq, $offset, 1;
   }
 print qq{$seq};
"
01234567890123456
atcgcgtacatcgatac
AtcgcgtaCatcgatAc
[download]

Easy to put that into a function.

Update: Looking back, I see that BrowserUk essentially gave this solution already, so... Never mind.