I'm 23, have a Computer Science degree, and have been working as a backend engineer for about eight months. My university education covered programming, algorithms, networking, databases, and other fundamentals, but it didn't teach me how a complete production system is built and operated. I had little experience connecting frontend code, APIs, services, databases, cloud infrastructure, CI/CD, logging, testing, and deployment before starting my job.
Now I work with Python, FastAPI, JavaScript, SQL, and Azure on software used by real customers. I investigate tickets, clarify requirements, navigate an existing codebase, change APIs and backend logic, occasionally modify frontend code, work with databases, open pull requests, respond to reviews, and communicate about releases. I know I'm contributing, but AI is involved in most of the technical process.
I often use AI to locate relevant code, explain architecture, decide where a change belongs, write implementations and tests, debug problems, and review solutions. I understand the result afterward, but without AI I'm afraid I would struggle to know where to begin. I don't yet have a strong instinct for tracing a request from the browser through an API and services to a database, or for forming a useful debugging hypothesis.
I also keep creating ambitious study plans covering JavaScript, SQL, Docker, Linux, system design, cloud, and testing. I start courses and projects enthusiastically, become overwhelmed by everything I still need to learn, abandon the plan, and eventually start another one. I want to improve, but I no longer trust my own learning process.
My job gives me considerable flexibility, so I have a good opportunity to improve while gaining production experience. I still want to use AI, but I don't want it to be the thing holding my entire engineering process together. My goal for the next year is simply to become an autonomous junior engineer who can investigate a backend ticket, decide where to look, form hypotheses, implement and test a change, debug it, and explain why it works.
What would you recommend to a junior in this situation? If you once felt that experienced engineers knew an impossible amount, when did that feeling begin to change?
5 Answers
Experienced engineers aren’t holding the whole internet in their heads. They’ve seen recurring patterns, know which questions to ask, and have learned how to narrow a problem. They still look things up and use tools constantly. Your goal isn’t to memorize everything or avoid AI; it’s to remain capable of checking its suggestions, spotting bad assumptions, and explaining the system yourself. That confidence usually grows gradually after enough cycles of investigating, shipping, and maintaining software.
AI should amplify your thinking, not replace it. Before asking for help, spend 20–30 minutes trying to locate the relevant code yourself. Find the route, controller, handler, or job mentioned by the ticket, then trace the flow through the service, database, and response. Write down a few possible causes or implementation locations. After that, ask AI to critique your plan or explain a specific piece instead of asking it to invent the entire solution. That repeated practice is what builds the internal map of a codebase.
You’re already doing real engineering; the part you’re missing is mostly navigation and debugging instinct. For every task, try writing a short investigation note before using AI: where the request enters the system, what components it touches, what you expect to happen, and three hypotheses for where the issue might be. Then use AI in small, targeted chunks. After the change is merged, summarize it in your own words. That loop will teach you much more than reading an AI-generated explanation after everything is finished.
Don’t turn your year into another giant curriculum. Pick one real problem at work or one small project and follow it through completely: understand the requirement, trace the existing behavior, make a design choice, implement it, test it, deploy it, observe it, and document what you learned. When you’re blocked, ask a senior or AI focused questions, but keep ownership of the reasoning. Consistent, deliberate practice will matter more than collecting courses or mastering every tool at once.
There isn’t a shortcut around experience. Before AI, engineers used books, documentation, search engines, and questions to other developers. Those tools can speed up learning, but they can’t give you the pattern recognition that comes from solving many different problems. Let yourself be a beginner and focus on understanding the code you ship well enough to explain its behavior, tradeoffs, tests, and failure modes.
A small end-to-end project can help too. Build something with a FastAPI backend, a simple page, a database, tests, and a deployment. Make the architecture decisions yourself and use AI mainly for boilerplate or clarification. Greenfield work feels confusing at first, but completing one modest system gives you a useful mental model for production work.