Skip to content
Live
  • Lagos --:-- WAT
  • Barça next: at Sabah FK · Wed 25 Nov, 18:45 WAT
  • Latest article: What it's like being the developer at a finance firm
  • Press ⌘K to search
  • Available for new projects
MOE#07
← All articlesPersonal · · 6 min read

What it's like being the developer at a finance firm

Fourteen months building banking portals and payment flows at WealthBridge. What money does to how you write code, and what nobody tells you.

In September 2025 I joined WealthBridge Financial Services in Lagos as a full-stack engineer and IT team lead. I thought I knew what the job would be: build things in Next.js, ship them, repeat.

Some days it is exactly that. Other days I'm building an internet banking portal before lunch and rescuing an executive's overflowing mailbox after it. Fourteen months in, here's what the job is actually like.

The job description is "whatever needs a developer"

In a finance firm, the person who "does tech" ends up owning a surprising amount. In one typical stretch I:

  • built a feature for the bank's corporate internet banking portal,
  • uploaded the weekly economic report the research team had just finished,
  • moved the company website's deploys from manual uploads to an automated pipeline,
  • fixed a careers page bug,
  • and helped someone whose email had stopped receiving because their mailbox was full.

At first that felt scattered. Now I think it's the best part. I see the whole business, not just a ticket queue. When finance explains how a payment really gets approved, I'm hearing it from the people who live it, not reading it in a spec.

Money changes how you write code

The biggest shift was learning that in fintech, the unhappy path is the product.

On a normal website, a double-click on "Submit" sends two contact forms. Mildly annoying. On a banking portal, a double-click on "Send" could move money twice. So every transfer carries an idempotency key, a unique ID created before the request leaves the browser. If the same request arrives twice, the server knows it's a repeat and processes it once.

Other lessons came from the business side, and some of them surprised me:

  • A PIN is never stored. Anywhere. Not in local storage, not "just for this session". It's typed, sent, and forgotten.
  • Slowing people down can be the right UX. On most sites, auto-submitting once the last digit is typed feels slick. For authorising a transfer, the CEO asked for an explicit "Proceed" button instead, so you get a moment to look at what you're approving. We kept auto-submit on the login token screen, where speed is what you want. Same component, two different answers, because the stakes are different.
  • Sessions expire, and forms should survive it. If you're logged out halfway through a transfer, the form comes back as you left it after you sign in again. The PIN step doesn't. You re-authorise from scratch.
  • Test without moving money. When the GMD asked for the transfer receipt to "roll down" out of the confirmation card, I needed to run that animation fifty times. So the dev build got a "simulate transfer" button. Nothing leaves the account, and the real flow stays untouched.

None of these are clever. They're the kind of boring that keeps a bank's customers safe.

The best feedback arrives before it's said

Our leadership has a sharp eye. Fonts get noticed: one week we went from Montserrat to Poppins because the CEO didn't like how the dashboard read. Spacing gets noticed too, on the laptop he actually uses, not the big monitor I design on. The brand colours are fixed by management, full stop, even when changing them would bump an accessibility score.

Early on I treated this as rework. Now I treat it as a skill: learning to predict the feedback. When I tighten a dashboard card, I already know the next comment will be about the empty space it leaves behind, so I fix that in the same pass. When I add a component, I ask whether it's new UI he hasn't approved, because if it is, it won't survive review.

The quickest fix is the one nobody has to ask for.

The infrastructure nobody sees

The group website used to be deployed by hand: upload files through cPanel, log into the server, run the build, restart the app, clear the cache. Every update was a small ritual with plenty of room for mistakes.

Now a merge to main triggers a GitHub Actions pipeline. It streams the build to the server over an SSH key that can only run one script, then builds, restarts, purges the cache and runs a health check. If anything fails, it rolls back automatically. There's a dry-run mode too, so the pipeline can be tested without touching the live site.

While fixing the site's speed I hit my favourite bug of the year: the server's CPU was too old to run sharp, the library Next.js uses to optimise images. Not an old version of sharp. Any version. So the images are now resized and converted to WebP in CI, before they ever reach the server. Between that, smarter caching and removing 440 KB of fonts we weren't using, the mobile PageSpeed score went from 74 to over 90.

Nobody outside IT will ever notice any of this. That's the point.

Legacy code is the job, not an obstacle

Finance runs on systems that have worked for years, and "working" beats "modern" every time. When I designed a new Node.js service for payment approvals, it had to sit next to an existing PHP system, not replace it. That's the strangler-fig pattern: build the new piece alongside the old one, move work across gradually, and never bet the business on a big rewrite.

I set myself one rule for that project: I don't change the existing system unless my service genuinely needs to talk to it. Partly that's good engineering. Partly it's accountability: if something breaks, I want to know for certain it wasn't me.

The other rule was to trust nothing that arrives from upstream. Before the service acts on an approved payment, it re-checks everything itself: the amount, the beneficiary, whether anything has changed since the approval. If Redis goes down mid-flight, a reconciler picks up where the queue left off. Most of the code is about what happens when things go wrong.

Understand it, or don't ship it

Midway through that build I hit a wall. Thirty files had changed, it worked end to end, and I couldn't fully explain how the stages connected. I could have kept going. Instead I stopped for a day and traced a single example payment through every file and every function until I could draw the whole flow from memory.

It felt slow at the time. It was the best day I spent on the project. I'm the one who'll be called when it breaks, and I'm the one who has to walk the next engineer through it. If I can't explain it, I don't own it yet.

Some wins have real people behind them

My favourite project here wasn't the most technical one. The Bridge Foundation's scholarship application system attracted more than 1,340 applicants and 50,000 page views in three months. During the application window I fixed problems as soon as they appeared, because behind every failed submission was a student with a deadline.

Later, a batch of assessment results ended up out of step with the records they belonged to. I recovered 1,345 rows in production with a careful three-way JOIN, and nothing was lost. Stories like that don't make a portfolio look flashy. They're what I'm proudest of.

If you're joining a finance firm as a developer

  1. Design for retries first. Idempotency keys, queues and audit trails aren't extras. They're the foundation.
  2. Talk to finance early. The real process is never quite the one in the document.
  3. Treat taste as a requirement. Learn what your stakeholders notice, and fix it before they have to.
  4. Automate your deploys before you need to. The first bad manual deploy costs more than the pipeline would have.
  5. Respect the legacy system. It's paying everyone's salary.
  6. Understand every line you ship. In fintech, "it works" isn't the same as "I know why it works".

Fourteen months in, I still get the mix of a startup and a bank in the same week, and I still go home to three kids who think my job is "typing". They're not entirely wrong.

If you're building something in fintech and want a developer who thinks about the unhappy path first, let's talk.

esc
Curriculum vitae

Moses Chukwudi Nwigberi · Résumé

The short version: WealthBridge, VisCorner, DITCOSoft, Pepnops and a few seasons of freelance.

Open the PDF