One regex approach might be:

var input = "hello world, h e l l o! hi. I am happy to see you!";
var output = input.replace(/(?<=\b\w)\s+(?=\w\b)/g, "");
console.log(output);

Here is an explanation of the regex pattern, which targets whitespace situated in between two standalone single word characters on both sides:

(?<=\b\w)  assert that what precedes is a single character
\s+        match one or more whitespace characters
(?=\w\b)   assert that what follows is a single character

As you have tagged pcre, another option could be:

(?:\b\w\b|\G(?!^))\K\h(?=\w\b)

Explanation

• (?: Non capture group for the alternation |
  • \b\w\b Match a single word char between word boundaries
  • | Or
  • \G(?!^) Assert the position at the end of the previous match, not at the start
• ) Close non capture group
• \K\h Forget what is matches so far, and match a single horizontal whitespace char
• (?=\w\b) Positive lookahead, assert a single word char followed by a wordboundary to the right

Regex demo

In the replacement use an empty string.

Another option could be matching at least 2 chars in a sequence of single word chars, and in the replacement remove the space.

Note that \s could also match a newline.

const regex = /\b\w(?: \w)+\b/g
const str = "hello world, h e l l o! hi. I am happy to see you!";
let res = str.replace(regex, m => m.replace(/ /g, ''));
console.log(res);