Oláyíwolá Odunsi

Oláyíwolá Odunsi · senior backend & cloud

I design & build infrastructure that scales.

Backend systems, platform engineering and cloud infrastructure — plus agent harnesses, the runtime scaffolding that lets LLM agents do real work reliably.

This is where I write about systems and anything that interests me.

trace / career

click a span to inspect

SPAN DETAIL

Senior Software Engineer · Openprovider

Oct 2023 → now · Rotterdam, remote

Core domain infrastructure as Go services on Kubernetes: pricing, availability, DNS, RDAP and authentication, holding a 99.99% uptime SLA across 4M+ domains. I own the architecture decisions the team builds on and mentor engineers on idiomatic Go.

Go · Kubernetes · MongoDB · Redis · MySQL · gRPC · OpenTelemetry

private projects / 3

01

HomeOS

Problem
Bookings and chef payouts have to settle correctly against inconsistent third-party payment webhooks, and the booking endpoint has to hold at peak load without dropping requests.
Decision
Built an idempotent webhook replay system that can reprocess the same booking any number of times without creating duplicates, and put a caching layer in front of the booking endpoint for peak load.
Result
Duplicate payouts at zero: a replayed webhook settles the same booking once.
0
Duplicates
Live product ↗ Go, Docker, Redis, Postgres, ElastiCache
HomeOS booking flow
02

Ìtàn

Problem
An LLM can describe an edit but cannot execute one, and a video agent that mutates state as it goes leaves nothing you can inspect, resume or trust twice.
Decision
Built from scratch in Go around a replayable edit ledger, where every agent action is an appended operation rather than a mutation, with ffmpeg and headless Chromium doing the work and models hot-swappable mid-session behind one interface.
Result
LLM-authored HTML/CSS/GSAP motion graphics render frame-accurate, and the same agent is drivable from a CLI, a browser UI or by voice.
Go, ffmpeg, headless Chromium, OpenRouter LLM APIs
Ìtàn editor mid-render
03

Khanzuo

Problem
Bug reports arrive as prose on tickets which makes reproducing them by hand slow.
Decision
An agent that drives the real app like a user, then hands back fix-ready context.
Result
Each report comes back as a replayed session with the failing step, console output and network trail attached, so the fix starts at the cause rather than the retelling.
Source ↗ LLM agent, Go
Khanzuo agent mid-run

log / writing

2026-08-05WARN
Your Go Service Is Getting OOMKilled and It Isn't a Leak
A highly concurrent Go service kept dying with exit code 137 in Kubernetes, but pprof showed no leak. The problem was that the Go runtime had no idea the pod had a memory limit. This is what GOMEMLIMIT fixes.
2023-08-04INFO
Create an API Rate Limiter in Golang
Learn how to create an API Rate Limiter in GoLang.
2023-07-11INFO
Production-Ready Beanstalkd with Laravel Queues
Queues are one of the integral parts of a scalable system, this article introduces the users to Queues and how they can integrate them into their Laravel application to improve the system performance.
2022-07-27INFO
Create a Crud Application With Golang and PostgreSQL
Learn how to create a RESTful API using GoLang and PostgreSQL.
2021-04-29INFO
Simple Way to Run Multiple Long-Running Processes in PHP Asynchronously
Learn how you can run multiple long-running processes in PHP asynchronously
2021-04-24INFO
How to Add Retries to External API Calls in PHP
When making HTTP API calls to external or third-party systems, failures may occur due to issues with the server connection or any intermediate hops in-between, this article explains how to implement custom Retry logic using the Exponential Back-off Algorithm.
2020-08-04INFO
Vue.JS Techniques to Save You Development Time
Vue.js have some intrinsic parts that are not widely known, and not even mentioned in the documentation, knowing some of these can save you time.

about

Oláyíwolá Odunsi

MSc Artificial Intelligence & Human Factors, Coventry. BSc Computer Science, Hertfordshire. Certified Kubernetes & Cloud Native Associate. United Kingdom, remote across GMT and CET hours.

I work on the unglamorous half of a product: the queue that has to drain, the endpoint that has to hold at peak, the migration that cannot lose a row.

Most of what I know came from things breaking in production and having to explain why. That's also most of what I write about.

Stack

  • Go
  • Kubernetes
  • MongoDB
  • Postgres
  • Redis
  • gRPC
  • AWS
  • OpenTelemetry

Also: PHP · Laravel · Vue · Node.js · RabbitMQ · Terraform · TypeScript