Shuga if we’ve met.
I build backend systems for fintech and AI. I’m mildly obsessed with ledgers reconciling on the first try, and I ship things faster than I probably should. That works out, because I’ve been doing this seven years and mostly know what I’m doing.
Last year I trained the computer-vision pipeline that got my sports-analytics startup a data-licensing deal with Nigeria’s top football league. Before that, I solo-built a Gen-AI fashion platform (mobile, web, admin) in under 90 days. Right now I’m Engineering Lead at OrdrGo, a social-commerce thing live in twelve markets, which mostly means I spend my life thinking about timezones.
Based in Lagos. Pair with people in Seattle, Sydney, and London. Answer emails same day, usually sooner.
Each card carries three hard numbers. If a project can’t produce three, it’s in “Also built” below instead.
I was the whole engineering team. Mobile, web, admin, and the Vertex AI try-on pipeline, all under one 90-day deadline.
I lead four people here. We turn WhatsApp, Instagram, and TikTok DMs into real orders. The hardest part isn't the messaging. It's the twelve timezones.
I co-founded this. I stand up the AWS GPU boxes, train the computer-vision models, and sit in the room when contracts get signed.
I built the entire backend, solo. Non-custodial crypto deposits, double-entry ledger, treasury spreads, and the PalmPay fiat off-ramp, all live at jexpay.io.
Built the core backend. WhatsApp-native agency banking: the kind of thing that competes directly with OPay and PalmPay, except field agents don't need hardware.
A virtual-account engine that's been quietly running every reconciliation path at the bank since 2022. Nobody pages me about it anymore, which is the point.
Core card-issuance rails, virtual and physical. Also built the microservice that cut dispute resolution from six hours to two. That was a good week.
Cross-border payments and multi-currency settlements for diaspora-led businesses. KYC/KYB and payouts, all wired up.
Brand-to-vendor marketplace. Around 2,000 verified service providers last I checked.
EdTech platform for Business Analysis and Data Analysis programs. Ran the engineering team that shipped it.
Corporate site and Android mobile app for a group holding company. Nothing fancy, just solid.
Donation and outreach platform for a non-profit working in education and healthcare.
Danish electronics-repair marketplace. First time I ever got paid to write code (2019).
Corporate site and internal portal for a Nigerian group of operating companies.
The Banking-as-a-Service API layer at Heritage. 150+ SME partners integrated over OAuth 2.0.
First internship. Ruby on Rails. First time anyone paid me to type.
Backend developer. Node.js and MongoDB. Also the first 3am page of my career, which was educational.
First fintech role. Learned that ledgers don't forgive and that's why they work. Built POOLER.
First US client. Card-issuance rails and the microservice that cut dispute resolution by two-thirds.
Lead Software Engineer. First team I ran: five engineers across a marketplace and an edtech platform.
Started training computer-vision models in the evenings. Ended up on a licensing call with a national football league.
Solo-built a Gen-AI fashion platform in under 90 days. Mobile, web, admin. Slept eventually.
Promoted to Engineering Lead. Four teammates, twelve markets, still answering emails same day.
I’m Abiodun Abdullahi, 29, from Lagos. The nickname stuck somewhere around secondary school. I stopped fighting it around the time I realised everyone was going to call me Shuga anyway.
I didn’t study computer science. I studied Mathematics Education at the University of Ilorin, which, if you’ve never had to do it, is the very specific discipline of explaining complicated things to people who don’t yet want to hear them. That turned out to be half this job. The other half is that the math itself still shows up: it’s why my ledgers reconcile on the first run, and it’s why when my co-founder and I needed computer-vision models at Roundup, I trained the YOLO and ResNet myself instead of outsourcing it. The distance between “I learned this in another language” and “I can ship this” was smaller than it looked.
I write on Medium. Lately I’ve been arguing that LLMs are not products, that your state-of-the-art model will lose to good data ops, and that latency beats accuracy in sports tech. It’s opinionated on purpose. Mostly I’m trying to save other people the six months I burned learning each thing.
Off the clock: football (watching more than playing — knees have opinions now), jollof that is slowly, annually, inching toward my aunt’s, The Alchemist reread more times than I’ll admit, and the kind of walks where the whole point is the walk.
A pushback on vibe-engineering. If your AI agent's whole architecture is a prompt, you don't have a product, you have a demo that crashes in production.
Feb 2026Nobody loses because their model was wrong. They lose because their data operations were bad and they couldn't see it. Notes from shipping ML at Roundup.
Nov 2025You can have fast or you can have perfect. At Roundup we had to pick. It wasn't the pick I expected to make.
Nov 2025Hiring for a senior backend or engineering lead role? Send a note. I read everything. I reply same day, usually. No forms, no filter, no “we’ll be in touch”.