The author seems to be misinformed about the relative value of compiler front-end versus compiler back-end.
Front-end is not where the challenges are in any production compiler. The hard work is in the optimizer. Lexical and syntactic analysis are homework problems for sophomores. That is not where any of the value in a production-worthy compiler lives, no more than the value of your C++ code being found in the semi-colons at the end of the statements.
Wrt to the outline for someone wanting to self study in the area, what are you thoughts?
By "this area", are you referring to front-end, or back-end?
I am going to guess back-end, because there are a lot of self-study tutorials for front-end that you probably have already found. Unfortunately, I am not an expert in back-end, although at one point I did manage a group that was involved in compiler validation -- but I was pointy-haired, it was my team that knew the innards of the compiler.
Advanced texts in compiler construction are going to get into data flow analysis and liveness testing, and talk about basics of code motion. These are all elementary topics and barely touch on the state-of-the-art, but are foundational. Also, get good at reading the assembly language for the machine of your choice and look at the .S files.
Sorry I can't be of more help, but maybe I gave you some search terms.
I think there are a couple things in play here. Folks working with text, semi-structured data, synthesizing from disparate sources, etc will be front end heavy. Tokenization, lexing, is important outside of more than compilers, like loading binary formats from network or disk into memory.
For backend work, being able to extend or modify existing backends is important for languages targeting different runtimes (Spark, Beam, Impala). This can be in targeting new architectures or for predicate pushdown into data processing pipelines. Lots of different applications to use those skills.
Compilers and Database systems are an incredible microcosm of many areas of CS.
Areas of self study I think are nice are
MAL - Make a Lisp https://github.com/kanaka/mal
Nand2Tetris, project 11 https://www.nand2tetris.org/project11 (one should start from zero and make your way here, it is journey not a destination)
An educational software system of a tiny self-compiling C compiler, a tiny self-executing RISC-V emulator, and a tiny self-hosting RISC-V hypervisor. http://selfie.cs.uni-salzburg.at
LLVM is a huge system, libFirm is a much smaller, simpler system that includes a c front end. From their site
> libFirm is a C library that provides a graph-based intermediate representation, optimizations, and assembly code generation suitable for use in compilers.