^ This. Also, honestly, I would suggest focusing on radical project management skills. Being an AI engineer isnt just about code, it is basically about managing expectations because stakeholders usually think AI is magic. You might want to consider learning how to say no when a project scope creeps into impossible territory. I remember back when I was starting out, I wasted months trying to squeeze accuracy out of a dataset that was fundamentally broken because I was too scared to tell my lead it wasnt possible. Be careful with that. You really need to develop a thick skin for iterative failure. If you cant communicate why a prototype is failing without making excuses, you wont last long. Make sure to keep your notes organized too. I personally found that using a Remarkable 2 Digital Paper Tablet 10.3 inch helps me map out these logical flows away from the screen, which keeps my head clear. Also, watch out for the cost of compute. You gotta be frugal with your cloud instances. If you dont track your spend, you will end up blowing your budget on a weekend test run. I keep a constant eye on my local development performance using a Dell UltraSharp U2723QE 27 inch 4K USB-C Hub Monitor just to ensure I am not wasting cycles on bad code while I am still drafting. Honestly, learning how to be efficient with your resources is just as vital as knowing your math.

Honestly, focus on domain expertise and communication. Being able to explain why a model matters to a business is huge.

• Problem translation: turning business issues into logic
• Ethical awareness: checking for bias before deploying I personally use a Logitech MX Master 3S Wireless Mouse because it helps me stay focused during long review sessions. It works well and I have no complaints regarding the reliability of my workflow now.

@Reply #1 - good point! Honestly, empathy and curiosity are way more important than people think. You gotta be able to translate human friction into data requirements. TL;DR: focus on story-telling and documentation. I actually keep my documentation clean using a Keychron K2 V2 Mechanical Keyboard Brown Switches to make sure I never miss a beat. Being able to explain your weird model architecture to a non-tech manager is an absolute superpower, trust me.

Omg I am in the exact same boat as you! I have been struggling with this exact pivot for like three months now and honestly it is so frustrating because everyone just tells you to learn more math. I feel like I am constantly hitting a wall and honestly I have no idea if I am even moving in the right direction or just wasting my time. It is super comforting to know someone else is dealing with this too, because it feels pretty isolating when you are just grinding away in Chicago trying to figure out what actually matters. Seriously, if you find any resources that actually help clear up the confusion, please share them because I am just as lost as you are!