Getting A Senior Backend Software Engineer Job in the AI Era

Disclaimer: None of this is professional advice. I'm not a professional. I'm just loud about my opinions and occasionally employed.

So you got the "we value your contribution" email.

I've worked at three start-ups that started-up, doing the right thing at the right time, but the right time passed, and every other company was also doing the same thing at the same time. Because I don't have million$ lying around to retire early, I had to scramble for new jobs, and in the process I've made a few notes worth sharing.

Before you even start applying

If redundancy is on the table, don't just take it. Ask if they'll do a pay cut or reduced hours instead, like a four-day week, while you look for something else. It beats watching your savings evaporate while doom-scrolling job boards. They might say no. Doesn't matter. Ask anyway.

And when the "here's your severance" conversation happens, don't just nod and sign. Have an actual conversation with a legal advisor about what you're entitled to before you agree to anything. You can't un-sign a thing.

The market has changed, and not in the fun way

There are many people looking for work right now, and companies know it. They can afford to be picky. I did roughly eight final-stage interviews before I got my current job. Eight. Not eight interviews total. I did eight final stages, the ones where you think you've basically got it.

Here's my working theory on why: I wasn't specialised. I bounced between different bits of tech over the years, so I applied for jobs where I wasn't quite the perfect fit, and companies picked the person whose CV matched the spec exactly. Generalists lose to specialists at the CV-screening stage, every time. You need to look like the answer to their exact question, not a good answer to a slightly different question.

There's an interesting quirk of modern job hunting that seems counter-intuitive. It's obvious that companies will reject you for being too junior, but they also reject you for being too senior, on the theory that you'll get bored and leave in six months. What that means is that, if you're desperate, you may apply for junior roles, with salaries below what you're worth. You're not going to get those... there are thousands applying for those. What companies are mostly looking for are people who can complete tasks and look after software by themselves. Aim for the job you're already doing, or perhaps with a bit more responsibility, even nudging your salary range slightly higher. Employers want people who are worth a lot.

And yes, they will ask "what's your AI workflow?" Many do this! Have a real answer ready. "I asked Claude Code to do it" is not a good enough answer, even if, behind the scenes, you're the Lord of the Bots. You need to be able to explain this well, along with the shortcomings of AI.

See my blog post on AI dev for more nonsense on this topic.

Use AI to prep

You can use a Claude project for this. The project has many files, but the most important is a big document of everything you've ever worked on. Markdown is probably best. Then write prompts that tell the AI exactly how you want it to answer the standard interview questions, using your history and the job spec together. The standard questions are always the same:

• "Tell me about yourself." They always ask this. Have a template that adapts per job, including a few lines establishing that you've done this exact job before, somewhere else, and you were good at it.
• 
  "Why do you want to work here?" Not salary, not location. This is about matching your expertise to what they do, and showing you've actually looked at what they do. Somewhere between "obsessed super-fan" and "didn't check the company name" is the target. It's really important to hit the sweet spot here - this was where I was constantly filtered out in the beginning because I was "not enthusiastic enough." When I showed super-fan level enthusiasm, I got the generic "Thank you for your..." When I hit the sweet spot, I started getting interviews. It doesn't actually matter that you're really enthusiastic - you need to explain your enthusiasm in the right way. Their website and job spec can help with that... there is often a careers section where they will tell you why you will love working there.
• Questions for the interviewer. You have to ask something, at both the hiring-manager stage and the final stage... not necessarily the questions that Claude tells you to ask, but something to show you've been listening. Interrupting the interviewer to ask them about something they've said, because you're genuinely interested, might be better than listing generic questions at the end. I got rejected after a 6 stage interview process because I "wasn't dynamic enough". Ask about the interview process itself while you're at it. Interviewers often give you hints about how to do well. This is gold - make notes! Especially for coding tests - you can actually ask what they're looking for, if they don't already tell you.
• Research the company first, before any of the above. What they do, what they claim to value. I once didn't get a job offer after doing well up to the final stage, because they asked a values-based question... "Are you the kind of person that invites people to the pub after work?" Had I known their core value was about being a friend, I could have responded, "No, I organize other socials" rather than just "No." If they do ask weird questions you can always ask them what they're hoping to learn from the answer, and then give a proper answer that actually answers the question.

Get the AI to do mock interviews too, ideally with voice mode. Practising thinking on your feet out loud is a completely different skill from writing good answers, and it's the one that actually gets tested live.

While I was doing all this, I got AI to quiz me in a way that it tracked what I knew, so I didn't have to answer the questions to topics I already knew well. I did this by building a website that does it properly: fullmarks.quest. You give it a topic, like a job spec, React, AWS, whatever's making you nervous, and it builds you a course, quizzes you on it, works out exactly where your gaps are, and then uses spaced repetition to make sure the gaps actually close instead of getting quizzed away and forgotten by Thursday.

Don't wait to be asked the right question. Interviewers work off an arbitrary checklist that may or may not cover the interesting parts of you. If there's something worth knowing that they haven't asked about, just tell them.

Your CV needs one thread, not several

The failure mode is a CV that's trying to be all things, like full stack, but backend-leaning, but a React specialist, but open to lead roles, but also individual contributor. Pick the story. Filename, title, and experience section should all point at the same job.

Some specific notes, on a real CV I was sent:

• If the file's called "Full Stack" but the title says "Backend," that's a fight your CV is having with itself before anyone's even opened it.
• Titles matter less than you'd think, honestly. I got hired as "senior backend" and my contract says Staff Software Engineer. Nobody's checking.
• Drop weakening words. "Strong React specialist". "Strong" undercuts "specialist." Either you're a specialist or you're not.
• Skip "15+ years." Just say the number. The plus is doing something I can't quite name, and I don't trust it.
• Bold your tech stack in the experience bullets. If someone's skimming, and they are always skimming, they should be able to see "TypeScript, three jobs" at a glance.
• Find your numbers. Cost cut, time saved, incidents reduced, anything. If you genuinely don't have a number, work backwards and make a reasonable estimate rather than leaving it blank.
• If your job history has a run of short stints, have a real answer ready for why, because they will ask, and "it just didn't work out" three times in a row is not an answer, it's a red flag wearing a trench coat.

If you're going for lead, prepare differently

Titles are mostly noise, like I said, call yourself senior, lead, staff, principal, whatever fits the spec, nobody's checking your payslip. But the interview for a lead role is a very different interview, and treating it like a slightly-more-senior version of the IC one is how you get caught out.

Almost regardless of the exact title, expect mentorship questions in senior roles. That one's universal. But for anything explicitly "Lead," get ready for a run of questions about how you actually work with other roles and steer a team:

"How did you work with the product owner?" "What's your leadership style?"

These aren't technical questions and you can't wing them on the day. I found them genuinely hard, and needed proper prep time before I was comfortable answering, not because the questions are clever, but because "leadership style" is the kind of thing you've never had to articulate out loud before, even if you've been doing it for years.

As a tech lead I used an approach I called, "The Definition of Awesome". That landed well. Maybe I'll do a blog about it one day.

The stuff you're expected to know cold, backend edition

If you're going for anything backend-leaning, here's the syllabus, apparently:

• 
  Observability. Name real tools your company used, like Sentry, Prometheus, Kibana, the ELK stack. Know when to log info versus warning versus error, and have an opinion on structured logging. For AWS specifically: X-Ray for tracing, CloudWatch for logs/metrics/alerts. For GCP: Cloud Trace. Know what RED stands for. If someone asks how you'd handle logging, don't say "with a logger". Say whether you use structured logging and what fields you included.
• REST status codes. All the ones that come up in real arguments: 200, 201, 401, 403, 404, 429, 500. Know when to retry or fail.
• Microservices, and specifically when you'd reach for a synchronous API call versus an async message-broker-and-handler setup. This comes up constantly.
• ACID and isolation levels. Fintechs in particular love to test you on "how do you make a safe payment or withdrawal," which is really just a transactions question.
• System design. This is the one that scared me most. Watch walkthroughs. This playlist got me through it. Practise a handful of designs until you're not improvising from zero, and make sure your DB / SQL choices are solid. You probably won't be asked to write SQL that actually works. Pseudocode is usually good enough, as long as you include your isolation levels and locking.

Or, you know, build a course on it in fullmarks.quest and let it figure out which bits you actually need to drill. It doesn't care that system design terrifies you. It just keeps asking until it stops terrifying you.

Cloud specifically, like AWS or GCP, is mostly a concepts test, not a "write me some Terraform live" test. Nobody's ever asked me low-level infra questions. Have at least one real story about something you actually deployed, and understand the concepts well enough that they transfer between providers, because they mostly do.

One more thing that took me too long to figure out: you don't need to remember the deep details of everything you built. "I don't remember the exact details, but if I were doing it today, here's how I'd approach it" is a completely fine answer. Seniority doesn't mean total recall. It means you know what a good approach looks like even when the specifics have faded.

Where to look

LinkedIn will happily let you apply to hundreds of jobs and reject you from all of them, usually with a template email that manages to say nothing at all. It's possible that you'll get an occasional interview, but it's a low-yield, thick-skin-required strategy, and writing a tailored cover letter for each one, then editing it so it doesn't read like Claude wrote it, gets old fast. After many depressing, "Thank you for your application..." emails, I stopped applying on LinkedIn.

Turning on "open to work" seems to get recruiters coming to you instead, often in a weekly rhythm. Tuesdays, in my experience, for whatever reason. Recruiters who reach out have effectively pre-filtered you against a real role, which is worth something.

My highest probability of success was via people I knew - either former colleagues or recruiters who had great experiences with me in the past. With former colleagues, you can get valuable information, like exactly what they're looking for in the coding test. Better to delay the interview and prepare properly than give up the rare opportunity by handing in a coding test that focused on robustness, when they just wanted performance.

A few other places that produced better results than LinkedIn, for what it's worth:

• TotalJobs - for regular email alerts on new postings
• CV-Library - I suspect recruiters source a lot of CVs from here
• The Work From Home Hub - for remote-specific roles
• Wellfound - startups (it probably helps to be the first to apply for a job that's just been posted)
• Cord and Welcome to the Jungle - companies reaching out to you directly
• Experis, ContractorUK, Jobsite - if you're going contract

Pace yourself with the applying, by the way. Blast out fifty in a weekend because you're anxious about the silence, and you'll have fifty responses to prepare for the following week, which is its own kind of overwhelming.

If you take one thing from this: figure out the two or three gaps that are actually costing you interviews. Use a spreadsheet, and track where things are going wrong and right, so you get a vague idea of how things are progressing. Solve any gaps properly instead of vaguely re-reading the same AWS docs page for the fourth time. The tool for that is fullmarks.quest. You get free credits, and if you burn through the credits, tell it I sent you.

Well Structured AI Development Practises

We've all tried AI by now.

Some promote a build as quickly as you can approach, while others look at me in horror, as if using AI extensively melts your brain (it does).

The question is whether you're getting consistent, trustworthy output (the kind you'd stake a production release on) or are confident-looking code that half-works and then breaks something at 4:59pm on a Friday?

There's a gap between teams getting serious results with AI and teams getting occasional lucky ones. The gap isn't which tools they've licensed. It's whether anyone has sat down and thought carefully about how those tools fit together, and made it someone's job to keep that thinking current.

Someone has to own this

Every org needs at least one person (a human person) who goes deep on AI-native development. Not "I've skimmed the documentation" but deep enough to write guidance that stops every other engineer from reinventing the wheel, to maintain the high level context files that teach our agents how the organisation actually works, and to have documented guidance and a considered answer when a new tool appears and everyone looks around waiting for someone else to evaluate it.

Without this person, we get well-intentioned chaos. Everyone experimenting independently. Nobody sharing what works. The wheel reinvented in every corner of the codebase. Coding agents that behave completely differently across repos because nobody's maintaining the context files and they've drifted into contradiction.

One responsible person, the go-to guy, fixes most of that.

Context is the actual product

Prompt engineering is what you do for a single conversation. Context engineering is the discipline of building the right information environment for every conversation. It's the difference between briefing a contractor well and just handing them a laptop and pointing at the codebase.

Imagine hiring a contractor who's genuinely skilled. Worked with a hundred teams, knows the patterns, no ramp-up required. But they've walked through the door cold. They don't know you're on Python 3.14. They don't know the auth layer is a shared service. They don't know you spent six months migrating away from the ORM they're about to suggest. That contractor will make expensive mistakes in the first week, not because they're bad at their job, but because they don't know your job.

Context files are the briefing.

Organised in a hierarchy: org-level at the top (your stack, your architecture principles, the things that don't change project to project), project-level below that, feature or component level below that for focused work. The AI lead owns the top. Everyone else contributes to the levels below.

Treat context as code. Version control it. Put it through pull requests. Iterate on it. Run a retro at the end of each sprint and ask: did the coding agents do anything surprising or wrong? Consider if the answer points to a gap or an error in a context file.

Fix it. Ship the change. Check again next sprint.

A fast workflow for big changes

The default way to work is: open chat window, describe the problem, accept or reject what comes back. That works for small, isolated tasks. See Managing risk at the bottom of the blog.

The following idea comes from https://tylerburleigh.com/blog/2026/02/22/. I don't have evidence that it works perfectly - just sharing what I've found. It's a staged approach with review cycles between stages. Errors caught at planning stage cost almost nothing. Errors caught after they've been faithfully implemented across multiple stages cost quite a lot more. So you front-load the thinking.

Research first. Before any code is written, the coding agent researches the codebase and / or the web, and writes its findings to RESEARCH.md. You read it. You correct any misunderstandings. Only when that's solid do you move on. Start a fresh session. Model performance degrades noticeably once you've used around 40% of its context capacity. Fresh sessions keep it sharp.

Then plan. The agent reads RESEARCH.md and produces PLAN.md: a breakdown of the work into stages, plus CHECKLIST.md. A separate session then reviews the plan from the perspective of a senior engineer. You read that review. Accept or override its recommendations. Revise until it's ready.

This is the highest-leverage point in the entire process. A bad assumption that survives planning gets implemented faithfully, and you'll discover it somewhere around stage 4, trying to explain why the feature is broken in a way that requires unwinding three phases to fix.

Then implement, one phase at a time. Each phase in its own session. Committed when clean. Reviewed before moving on. Use different models for implementation and review. A model that just wrote the code is not ideal to find problems with it. It'll tell you it's fine.

It's not always fine.

LSP plugins give the coding agent live type information from the IDE, so it catches type errors and missing imports without waiting for the compiler. 

Consider Git worktrees if you're running parallel features: separate agents, separate branches, separate directories, no collision, merge as normal.

When it's running well, consider automating the middle

Once this is working reliably you can remove human involvement from the implementation loop. Not from the edges, but only the middle.

The human still approves the plan before anything is built. The human still reviews the PR at the end. What changes is what happens between those two points: an implementer agent executes against PLAN.md, automated gates run (tests, lint, type checks), a reviewer agent evaluates against the plan, and they iterate until clean. No human required in that loop.

The PR is self-documenting. RESEARCH.md, PLAN.md, and the commit history are the description. The engineer can see exactly what the agent understood, what it planned, and what it did.

Nothing's a black box, because "the AI did it" (which is as tempting to say as eating a warm, golden, honey-glazed doughnut oozing with rich, velvety Marmite custard) is not an acceptable answer in a code review. Not yet, anyway.

Managing risk

Not every change carries the same risk, and it helps to think about this on two dimensions: likelihood of something going wrong, and impact if it does. For likelihood, ask how complex the change is, how well-tested the affected code paths are, and how reversible the outcome is. For impact, ask how many users or services are affected and whether you can recover quickly if something breaks. A rough mental multiplication of the two tells you how much caution is warranted.

When I work in core services in financial applications, I micromanage Claude. What do you do?

TDD, or should Claude just write everything?

Let's start with a question you've been avoiding: do you actually do TDD, or do you nod along at the conference talk, then go home and write the implementation first like everyone else? Be honest. The tests aren't listening. The uncomfortable truth is that TDD has always demanded a very particular flavour of masochism: sit down, write code that deliberately fails, resist every instinct to just make the thing work, and trust that the scaffolding of failure will eventually produce something beautiful. Most developers respect this in the same way they respect flossing. And now here comes Claude, ready to help. What could go wrong?

Quite a lot, as it happens, and in a rather specific way. Left to its own devices, Claude will skip the red phase entirely. It's too helpful. You ask it to write a test for a feature that doesn't exist yet, and, like an eager intern who's read ahead, it writes the test and the implementation in the same breath, hands them to you simultaneously, and waits for praise. The tests pass, obviously, because Claude just wrote both sides of the conversation. This is the AI equivalent of marking your own homework, then putting a gold star on it. SD Times put it this way: AI doesn't eliminate TDD, it exposes whether you understand it. Which is a polite way of saying it will cheerfully help you do it wrong if you let it.

The solution is that TDD and AI are actually a spectacular match, provided you hold the leash. The cognitive barrier that kills TDD in practice is that writing a test requires defining the interface, which requires understanding the architecture, which requires mentally sketching the implementation anyway; the circle eats itself before you've typed a single assertion. AI dissolves that barrier completely. Let Claude draft the test, you review it, then let Claude implement against it, one cycle at a time, not in bulk. 

In other words, use this exact prompt:

"Sit. Good boy. Now, write one failing test for the login function. Just one. Don't implement anything. Don't even think about the implementation. I can see your little cursor twitching. Stop that. Write the test, run it, show me it's red, and then sit back down. If you've written only the test and nothing else, you'll get a biscuit."

Or, if you're feeling boring, something like this:

Do TDD, one test at a time, like this:

    - Prefer doing the happy path test(s) first.

      When testing edge cases, test as close to the edge as reasonable.

    - Implement code structures, so types used in tests work, but

implementation does nothing or raises, e.g. a method that raises

an exception

    - Write the test

    - Check it fails

    - Implement the literal minimum fix here.

      If more would be done in a full implementation,

      write a one line TODO in the implementation.    

    - Check it passes

    - Refactor if necessary

Now, some of you will have noticed a tension and are feeling very pleased with yourselves about it. If we're telling Claude to stop after every single test like an overexcited Labrador, when exactly does the plan happen? Surely we want Claude to think before it acts? Yes, probably. The thing is, plan mode and TDD aren't quite solving the same problem (or at least, that's how it seems to me). Plan mode is about what you're building. TDD is about whether you actually built it. One is a map. The other is the slightly neurotic habit of checking you're still on the road every five minutes, which feels excessive right up until the moment you aren't.

The temptation is treating a good plan as a green light to let Claude implement the whole thing in one uninterrupted sprint. My suspicion is that this works right up until it doesn't, and when it doesn't, you're several pull requests deep into something that's coherent, confident, and subtly wrong in ways that are deeply tedious to unpick. The plan tells Claude where it's going. TDD, is just the thing that keeps it from lying to you about whether it got there. You probably want both: plan mode to decide what you're building, TDD to keep Claude honest while it builds it. But I'll admit I'm still working out exactly where the seams should be, and anyone who tells you they've fully figured this out is either very clever or writing a different blog.

---