Staff full stack (lean front-end) -> AI ... How?
Posted by no-bs-silver@reddit | ExperiencedDevs | View on Reddit | 7 comments
Hoping to hear from any one who was able to make the transition.
I am 10+ YOE in full stack, east coast,. Currently working remote in a larger but not "FAANG" tech company basically everyone has heard of/knows. Decent comp. But a few years back took a position on a more front-end concentrated team and just after it felt like AI stuff took off and now I feel regret. In my head it seemed like a slight risk at the time to lean a bit more on the front-end side but in this market now it feels pretty much like it was not a great choice.
I genuinely am interested in building things with AI. I am one of those perhaps crazy people who thinks it is going to fundamentally change the world. I would like to dive deeper and be building it as my day job and since I am not it also feels like I am getting left further behind besides just not getting to work on it.
I have been working on AI side projects but I guess nothing too extreme. Doing a udemy course. Working on an MCP server at work.
I know I am not alone but getting basically no traction when I apply anywhere. I am scratching my head because the vast majority of the job postings are asking for people who have been building AI systems for 2-4+ years already.... which to me seems like an incredibly small amount of people worldwide no? And they seem to all pull in different directions of AI (RAG, MCP, agent training, cloud service hookup, etc.) that I also get a bit confused on where my best use of limited free-time to study is.
I'm willing to take a paycut to an extent and go back to commuting so I don't feel particularly picky.
Any thoughts?
ninetofivedev@reddit
Easiest way to learn ai is to just setup Claude code or codex or cursor and start building some shit.
I prefer Claude code. Never leave the terminal. Combine ancillary knowledge from other sources with the AI tool itself.
One thing people, especially this sub, don’t give Ai credit for is that it’s actually a really useful tool for incrementally and precisely explaining something to you.
Overtime, you’ll figure out how to be more and more productive.
As for building Ai products, there is basically the LLM training side of things, which requires a pretty deep understanding of LLMs in general.
And then there is prompt engineering. You call an api like bedrock and basically just wrap your software up in their model.
juggernawddy@reddit
I guess im the opposite. I don’t see learning as an easy thing. I don’t think there are any shortcuts. Helpful tools sure, but etching deep understanding of something into my brain must be done by force. Ever code in pencil? Seems silly, but give it a shot, might be surprised how much you don’t actually know or understand something into
Idea-Aggressive@reddit
Are you sure you’re giving correct advice to this person? E.g. RAG is irrelevant today.
Gloomy_Cicada1424@reddit
Frame yourself as a strong product engineer adding AI features, not an AI researcher. Build 2-3 complete demos with RAG/evals/deploy, then make clean case-study pages. I’ve used Runable for that part because nobody reads messy repos deeply.
itix@reddit
2-4+ years of experience in AI sounds insane. It only got good recently, and I have been using AI for only 1-2 months. And I am on the AI front lines at our company. Idk, I feel we here in Europe are always lagging in new tech.
The AI topic is very broad. I have already built my own RAG, MCP server, trained an LLM (it was very bad), now using Codex as my coding assistant. Only Codex was important, while working on RAG was very educational. Get Codex or Claude Code, let it help on your personal projects.
khuskii@reddit
Is there room in your company to think of more universal components that could be adopted by other teams? Or patterns that could influence your mcps? I’ve noticed that there’s still so much room to standardize and organize skills and processes, and while Cursor and Claude rely heavily on React docs/more available open source devrel, it’s worth making sure that the company is unified in it’s approach to architecture, patterns, etc. Especially since there’s so many layoffs/pips and people getting shuffled, it is nice to be able to contribute immediately with little to no ramp.
uniquesnowflake8@reddit
I think you might need to really study and implement fundamentals like the underlying math, neural networks, learning and AI algorithms, and make sure you fully understand the fuzzy areas like RAG