Olaiya Joseph Oniya — Production-grade systems, in days.

Production grade. In 48 hours.

I build the systems other engineers say take months. Real-time mobility platforms. Distributed backends. Production Flutter apps. LLM proxies. Whatever your fastest engineer would ship, but shipped now.

Building since 2019

Production systems 9 services live

Lines shipped, '26 ~150,000

The work that wins doesn't look like the work other people ship.

Most engineers ship features. I ship systems. The TagRides backend isn't a Node monolith with a Postgres on the side — it's 9 Go microservices, MQTT over EMQX for real-time location, Kafka for event streaming, Kong as the gateway, PostGIS for routing, Tile38 for geospatial lookups. All of it dockerised, observable, replayable, and built solo.

That kind of depth doesn't come from following tutorials. It comes from owning every layer — auth, routing, payments, push notifications, the matching algorithm, the simulator that lets you test it without 200 drivers on the road. When I tell a client I can ship their MVP in two weeks, what I mean is: I've already built the harder version of their problem. Yours is a remix.

I'm Oniya. I build the whole thing — so you only have to hire one person.

Most clients reach out thinking they need a team. They need a backend engineer, a mobile dev, a DevOps person, someone for infra, someone for the marketing site. Then they spend two months coordinating four contractors and shipping nothing.

I do all of it. Backend in Go. Mobile in Flutter. Web in Next.js. Infra on Docker + Nginx + Postgres. AI integrations with LiteLLM and Foundry. Brand systems, design tokens, landing pages, contracts. One person, one timeline, one invoice.

That's not because I'm cheaper than a team. It's because shipping fast requires somebody who can hold the entire system in their head at once. That's the part that's actually rare — and that's the part I do.

The pact

↳ A first working version of your product, on a real URL, within 7 days of kick-off.
↳ Daily progress updates. No "I'll have something next week."
↳ A free 20-minute intro call before any commitment. If we're not a fit, we're not a fit.

02 Selected work

Four different kinds of hard problem.

Distributed systems. Native mobile. Editorial web. Production AI infrastructure. Each shipped end-to-end.

/ 01 2025–26

TagRides — the platform.

A production-grade ride-sharing backend for Lagos. Built solo.

Nine Go microservices behind a Kong gateway. Real-time driver and rider location pipelines over MQTT (EMQX) and Kafka. PostGIS routing with OSRM and Valhalla integration. Tile38 for geospatial proximity. Paystack payments. Twilio + Mailgun for comms. Firebase auth. A simulator service that lets you stress the matching algorithm without putting a single car on Lekki-Epe.

Go 1.25
Fiber
GORM
Kafka
MQTT / EMQX
Kong
PostGIS
Tile38
Valhalla
OSRM
Docker Compose
gRPC

9 services · 1 engineer · 1 city

/ 02 2025–26

TagRides — the app.

Native Flutter, web-first, clean architecture, real-time map.

Domain / infrastructure / application / presentation — four-layer architecture across 575 Dart files and 38 screens. BLoC state management. MQTT5 streaming for live driver positions. Google Maps + custom marker pipelines. Geolocator, secure storage, Paystack SDK, Pinput OTP, Firebase auth, Google Sign-In with web platform-view shim. Ships to Android, iOS, and Web from a single codebase.

Flutter 3.8+
Dart
BLoC
MQTT5
Google Maps
Firebase Auth
Paystack
Dio
go_router
Lottie
Web platform views

575 Dart files · 38 screens · 1 codebase

/ 03 2026

LLM proxy — production in 48 hours.

A multi-tenant LLM gateway with virtual keys, per-user budgets, and TLS-terminated public ingress.

LiteLLM in Docker behind nginx with Let's Encrypt automation. Per-buyer virtual keys (LiteLLM) with model allowlists and daily-resetting budgets. Postgres-backed spend tracking. Stripped client beta headers at the nginx layer to keep upstream Azure AI Foundry happy. From zero to public HTTPS endpoint ai.tagrider.com with two paying-prospect keys live — 48 hours of work.

LiteLLM
Docker Compose
Postgres 16
Nginx
Let's Encrypt / certbot
Cloudflare DNS
Azure AI Foundry

$ curl https://ai.tagrider.com/v1/messages \
  -H "Authorization: Bearer sk-..." \
  -d '{"model":"gpt-5.5", ...}'
HTTP 200 ✓

48 hours · zero to production · two buyers

/ 04 2026

tagrider.com — landing.

Editorial marketing site for the TagRides product. Brand system from scratch.

Next.js 16 + Tailwind 4 + Motion. A complete brand system — mark, lockup, colour tokens, type scale — codified in BRAND.md and reproduced verbatim from the Flutter theme so the marketing site and app feel like one product. Departure-board ticker for driver cards. Animated route diagrams. Lighthouse 90+ on mobile.

Next.js 16
React 19
Tailwind 4
Motion v12
Lottie
ECharts-GL
Cobe

Brand system → product surface, in days

Languages I think in. Tools I reach for.

Languages

Go (primary backend)
Dart / Flutter
TypeScript / JavaScript
Python (data, tooling)
SQL (heavy)
Bash

Backend / infra

Fiber, gRPC, REST
Kafka, RabbitMQ, MQTT (EMQX)
Postgres, PostGIS, GORM, pgx
Redis, Tile38
Kong gateway
Docker Compose, systemd
Nginx, Let's Encrypt

Mobile / web

Flutter (clean architecture, BLoC)
Next.js 14 / 15 / 16
React 19
Tailwind 3 / 4
Motion / Framer Motion
Three.js, GSAP, Lenis

Integrations

Paystack, Stripe
Twilio, Mailgun, Firebase
Google Maps, OSRM, Valhalla
Cloudinary
OpenAI / Anthropic / Azure AI Foundry
LiteLLM

Production grade. In 48 hours.

The work that wins doesn't look like the work other people ship.

I'm Oniya. I build the whole thing — so you only have to hire one person.