Honestly, go for the hardcore deep learning route if you want that 160k mark! I remember back in 2019 when I was sweating over my first neural net... seeing that loss curve drop for the first time was such a rush! DL is definitely not dying to automation. It is the engine that runs everything. Prompt engineering is super fun and I use it daily for RAG pipelines, but the high-paying Seattle roles want people who actually understand the math and PyTorch architecture. Plus, you can learn this stuff without going broke. I built my career on a tight budget! I hunted for a used NVIDIA GeForce RTX 3060 12GB GDDR6 on Craigslist just to run local experiments and it was a total game changer. If you want to save cash while you learn, just grab a Google Colab Pro Subscription for a month or two. It gives you amazing compute for peanuts compared to building a massive rig. Stick with the hardcore stuff and you will crush those interviews... it is way more sustainable than just chasing the prompt hype!

To add to the point above: I really tried to make prompt engineering my whole focus last year but unfortunately, it just wasnt as robust as I expected. I spent weeks fine-tuning system prompts for a client only for the next model iteration to completely ignore my constraints... it was super frustrating. Real engineering happens in the retrieval layer and the data pipeline, not just the chat box. If you want that 160k, you gotta look at how the data actually moves. I had issues trying to run everything in the cloud because of latency, so I ended up getting a NVIDIA GeForce RTX 4090 24GB GDDR6X just to prototype RAG pipelines without the API lag. Understanding how to optimize vector search and chunking is way more valuable than prompting. Prompting is basically just surface level. Stick to the heavy lifting, honestly... its where the job security is.

tbh I just saw this and had to jump in because the Seattle grind is real right now. I tried going down the prompt route with a few startups last year and it was honestly such a letdown... the pay ceiling is lower than people realize. Unfortunately, the job security just isnt there when a new model update can wipe out your whole workflow overnight. Wait, I gotta ask tho—what has your Python experience actually looked like lately? Are you doing heavy backend stuff or more like simple scripting? It makes a big difference for the pivot. If you want the big money, just stick with anything from NVIDIA Graphics Cards for your local setup. Honestly, you cant go wrong with AWS Cloud Services for the cloud side either. Just learn the infrastructure... prompt engineering is basically just a layer of paint.