##PowerShell, 162 bytes

function f{param($f)-join([char[]](65..90)+(0..5))[[convert]::ToInt32(-join($f|%{+($_-cmatch'[A-Z]')}),2)]}
($a=$args[0])+(f $a[4..0])+(f $a[9..5])+(f $a[14..10])

OK, a lot of neat stuff happening in this one. I'll start with the second line.

We take input as a string via $args[0] and set it to $a for use later. This is encapsulated in () so it's executed and the result returned (i.e., $a) so we can immediately string-concatenate it with the results of three function calls (f ...). Each function call passes as an argument the input string indexed in reverse order chunks as a char-array -- meaning, for the example input, $a[4..0] will equal @('0','E','R','0','a') with each entry as a char, not a string.

Now to the function, where the real meat of the program is. We take input as $f, but it's only used way toward the end, so let's focus there, first. Since it's passed as a char-array (thanks to our previous indexing), we can immediately pipe it into a loop with $f|%{...}. Inside the loop, we take each character and perform a case-sensitive regex match with -cmatch which will result in true/false if it's uppercase/otherwise. We cast that as an integer with the encapsulating +(), then that array of 1's and 0's is -joined to form a string. That is then passed as the first parameter in the .NET [convert]::ToInt32() call to change the binary (base 2) into decimal. We use that resultant decimal number to index into a string (-join(...)[...]). The string is first formulated as a range (65..90) that's cast as a char-array, then concatenated with the range (0..5) (i.e., the string is "ABCDEFGHIJKLMNOPQRSTUVWXYZ012345"). All of that is to return the appropriate character from the string.

PowerShell, 162 bytes

function f{param($f)-join([char[]](65..90)+(0..5))[[convert]::ToInt32(-join($f|%{+($_-cmatch'[A-Z]')}),2)]}
($a=$args[0])+(f $a[4..0])+(f $a[9..5])+(f $a[14..10])

OK, a lot of neat stuff happening in this one. I'll start with the second line.

We take input as a string via $args[0] and set it to $a for use later. This is encapsulated in () so it's executed and the result returned (i.e., $a) so we can immediately string-concatenate it with the results of three function calls (f ...). Each function call passes as an argument the input string indexed in reverse order chunks as a char-array -- meaning, for the example input, $a[4..0] will equal @('0','E','R','0','a') with each entry as a char, not a string.

Now to the function, where the real meat of the program is. We take input as $f, but it's only used way toward the end, so let's focus there, first. Since it's passed as a char-array (thanks to our previous indexing), we can immediately pipe it into a loop with $f|%{...}. Inside the loop, we take each character and perform a case-sensitive regex match with -cmatch which will result in true/false if it's uppercase/otherwise. We cast that as an integer with the encapsulating +(), then that array of 1's and 0's is -joined to form a string. That is then passed as the first parameter in the .NET [convert]::ToInt32() call to change the binary (base 2) into decimal. We use that resultant decimal number to index into a string (-join(...)[...]). The string is first formulated as a range (65..90) that's cast as a char-array, then concatenated with the range (0..5) (i.e., the string is "ABCDEFGHIJKLMNOPQRSTUVWXYZ012345"). All of that is to return the appropriate character from the string.