Engineer with 7 Years Experience - Best Way to Learn Beyond Courses & Books?
Posted by Nithish97@reddit | learnprogramming | View on Reddit | 9 comments
I’m an engineer with \~7 years of experience (around 70% frontend, 30% backend) and I’m getting into AI. Looking for clarity on the most effective way to learn it.
Long Udemy-style courses don’t seem to work well for me, and juggling multiple books feels scattered. I’m also not very interested in building small demo apps — I’d rather focus on understanding how to build real systems, automate workflows, and use AI to its full potential.
I’m trying to find a learning approach that’s structured, practical, and actually leads to real capability.
For engineers who’ve gone down this path:
- What learning method worked best for you?
- Did you stick to one source or combine multiple?
- How did you transition from theory to building real systems?
- What made things finally click for you?
Would really appreciate practical insights and experiences.
theofficialnar@reddit
You could’ve just asked an AI to provide you with a plan and get it done with.
JohnBrownsErection@reddit
So based on your added context(on mobile so doing my best on the formatting):
Pick one annoying workflow in your actual life/job. Something like:
“Take messy customer/internal docs, answer questions from them, summarize changes, draft responses, and log the result somewhere.”
Now build the stupidest version possible:
Call an LLM API.
Feed it some text.
Get an answer.
Watch it hallucinate like a Victorian child eating wallpaper paste.
Add retrieval.
Watch retrieval fail.
Add chunking.
Watch chunking fail.
Add citations.
Watch citations fail.
Add evals.
Realize evals are the adult supervision you were avoiding.
Add logging.
Add retries.
Add guardrails.
Add human review.
Congratulations, you are now doing “AI engineering” instead of tutorial necromancy. The thing that makes it click is realizing AI apps are less about “prompt harder, wizard” and more normal software systems with a drunk oracle bolted into the middle.
JohnBrownsErection@reddit
My own background is as a data science student so my own experience with machine learning and AI are fairly basic. I did code a brute force reinforcement learning algo for a project but that's not really useful for our purposes here. I'm still in the "learning programming and stats" stage of things.
Nithish97@reddit (OP)
Thanks a lot for your effort - really appreciate it!
generic_excuse_x@reddit
7 years experience and you want somebody to hold your hand while you learn ai agents?
And they say nobody wants to hire juniors anymore...
Nithish97@reddit (OP)
I never asked for hand-holding. I'm asking what approach has worked.
Earlier on, during college, I taught myself programming and built small apps.
When I started working, I picked up better practises, understood systems, and shifted my thought process to write better code. This happened as I gained exposure.
In my current environment, I lack that kind of exposure; it’s all fairly surface-level. Hence, I wanted to reach out to an audience who've already got their hands dirty and could share their two cents on it.
aqua_regis@reddit
You need to start small and simple and gradually increase. You can't go 0 to 100 in a single step.
JohnBrownsErection@reddit
My thoughts as well. Getting into AI like... developing LLMs? Using LLMs as part of a workflow? Working with big data and doing machine learning?
I hate the current trend of just slapping AI on anything without context.
Nithish97@reddit (OP)
My bad, I should have provided a bit of context, as AI is pretty much an umbrella term.
To clarify my focus: I’m not trying to learn AI at the core level (like training new models or doing deep research). I want to learn how to effectively use existing models to build useful systems — things like automating workflows, generating code, building RAG pipelines, and experimenting with agentic AI setups.