Endgame grind against a London player

The first deviation from a normal London game started at move 5 with 5... Nh5 which destabilized the black square bishop of white on f4.

To be honest, I don't know much about the theory after 6.Bg5 and according to the database and the engine, the best move is f6 which has a line of f6 Bh4 g6 (not g5 which is the only move that I calculated), g6 immediately imbalances the game which allows for some dynamic play from both sides, it also let the h5 knight back to the center via g7. Another viable move is Qb3 which is the typical queen move in the London opening attacking the vulnerable b2 pawn.
Anyway, in the game, I played 6.Be7 and traded the bishops, and allowed white to take over the center with Ne5 or Bb5 destabilizing my knight on c6. After Bb5, the best move is Bd7 protecting and preparing to take back with the bishop because in the game after I castle he took the pawn and immediately go for my weakness on c6 with Ne5 and also open a discovered attack on my h5 knight. The position was much more complicated than I thought because I don't lose the c6 right away after Nf6 because of Qb7 and after Qa4 I had Nd7 with various compensation for the pawn.

I made my only blunder under white's overwhelming pressure here with 11...Qg5??, effectively trapped my knight after 12.g4 for 12...Nf6 ran into Nf3 trapping my queen. Of course, I should study my tactics more diligently next to avoid stepping on my foot like this, and another part is that I don't know the opening that well and was quickly surprised by the imminent collapse of the position after the opening.
Luckily for me, my opponent didn't find the killer tactics for 2 moves straight and then blunder the win away with a greedy capture on c6 which is expected of play at this level. The best move after 13...cxd4 of course is 14.cxd4 kept the pawn structure in tact and kept the pressure on c6. Black can protect the c6 afterward but the black position will suffer from a lack of space and will soon crumble under white's precise positional play. Black is needlessly greedy with Nxc6 trading the good knight for the bad bishop and at the same time allowing dxe3 to damage the pawn structure which later proved to be fatal.
After 12.0-0, I'm still oblivious to the tactics and went Bb7 which is another inaccuracy because the d7 square is now no longer controlled by black, and Nd7 will either take the exchange or dominate with an outpost on c5. White responded with Qa4 - an inaccuracy and he can't trap my knight anymore. Although my pieces are misplaced enough so white still retained a sizable advantage.
The few moves following white's mistake are fairly standard with me rearranging my pieces to better squares on the queenside. The queen trade is slightly inaccurate according to the engine but I think that with fewer pieces on the board then I can stretch white's defense of the weak pawns a little bit easier.

23...Rab8 is another inaccuracy according to the engine, and it wants me to go a5 restricting the movement of the white rook on the fourth rank, I think I need to study my endgame more for me to understand the intricacy of these finesses. Anyway, I then go for the f6 Kf7 waiting move and make the king a bit more active. White finally made an inaccuracy which is a4, now both the c and b pawns are backward which made them more vulnerable to the rooks. I played the best move a5 taking control of the b4 square away from White. Subsequently, white made 2 mistakes in a row with Rh4 and e4, I think they are weak moves because it only helps Black to strengthen their grip on the position and they just cut the rook from the battlefield happening at the queenside.
I responded with a nature 30...d4 targeting the pinned c3 pawn but apparently, a better line is g5 deflecting the rook from the defense of the pawn, and then after Rxh6 Kg7 Rh3 dxe4 Black has a massive center which can be used to mobilize the pieces to better square and suffocate White. White is no engine though so he continues to make worse move after worse move with 33.Nc1? being the final nail in the coffin. A takeaway is that 31...Nc7 is a bit more precise but I was concerned about the complication after 32.e5 with white sacking the pawn for counterplay on Black's king. Taking the time to calculate better moves in a classical game has to be a number priority instead of making a safe move and missing a win.

White's position is now hopeless with one of the pawns on e4 or b3 falling. And with the marching of the connected passed pawns 36...e4 37.Rf5 d3, white cannot stop the pawns from promoting without a serious loss of material and then resigned. I think I need to analyze my lost games too but I can't help to stroke my ego after a win like this once in a while.