Skip to content
Jacob NolletteCloud & software engineer
Case studies

The work.

Long-form write-ups of real systems: what the problem was, which turns were wrong, what finally fixed it, and what it changed. Filter by topic.

Client work · Networking · Reliability

The Routing Flag That Took Five Sites Down

A VPN option swallowed the container network on a shared web server. Five sites went down for twenty hours, and the guard that should have prevented it had been silently dead for two months.

Infrastructure · Reliability

When the Failure Domain Was a Lie

Five storage pools were told to keep three copies of every write. They were keeping all three on the same machine.

Infrastructure · Reliability

Every Disk in the Organisation, on One Dashboard

Fifty-eight drives across four sites, one collector — written because the standard one reports no age at all for the drives that matter most.

Infrastructure · Reliability · Self-hosting

Every Log in One Place

Container logs, hypervisor syslog and firewall events from three sites, in one lake — and the sizing decision that keeps the loud feed from erasing the quiet one.

Infrastructure · Networking

Three Clusters, One of Them on Cellular

Two remote sites got their own single-node Kubernetes clusters, managed from the main site — and the one on a cellular link taught me to count bytes again.

Networking

A Network Where People Cannot Reach Infrastructure

Three sites, a software-defined mesh, and a zone policy that removed a rule with 6.3 million hits on it.

Reliability

Backups Nobody Had Ever Restored

Every database now dumps itself with its own tools, ships two directions, and gets restored on a schedule — because the only backup that counts is one you have put back.

Agents & AI · Self-hosting

Six Agents, One Channel Each

A fleet of specialist AI agents that could not talk to each other, and the database-level fence that finally let them.

Infrastructure · Reliability

Three Ways Unlocked State Ate Live Infrastructure

The same missing safety mechanism destroyed live resources three separate times. Each time it looked like a different bug.

Infrastructure · Self-hosting

Twenty Services, Self-Hosted

Ticketing, file sharing, meetings, search, backups, security monitoring and a container registry — replaced with software I own, in about six weeks.

Agents & AI · Infrastructure · Self-hosting

A Browser Farm for Agents

Ten Chrome browsers running in Kubernetes, each with a real profile and a real login, so an AI agent can use the web the way a person does.

Infrastructure · Reliability

Nine Copies of Every Write

My Kubernetes control plane had been running five times slower than etcd considers healthy for over a year. I spent two hours blaming Ceph recovery throttles, caused an outage proving it, and then found…

Agents & AI

The Question That Vanished, and the Fix That Was Almost Right

A self-hosted AI agent crashed with "no user query found in messages" — on a request that demonstrably contained a user query. The cause was an upstream renderer that validates messages after the server…

Agents & AI

Working Inside a Small Context Window

An agent went quiet mid-conversation. The documented threshold was wrong, the intuitive fix was backwards, and the real constraint was a number nobody had added up.

Agents & AI · Infrastructure

Two Cards, One Model, and the Ethernet Between Them

Pooling two Tesla T40s under llama.cpp's RPC backend bought a quarter-million-token context window on hardware I own — and cost half the generation speed, because every single token had to round-trip over plain TCP…

Agents & AI · Reliability

Turning a Document Archive into a Tool an AI Agent Can Use, Safely

A self-hosted document archive holds years of receipts, statements, and scans — useless to an AI assistant unless it can be reached safely from the outside. The build: an OAuth-gated bridge that took a…

Agents & AI

A Local Voice Agent, and the Two Weeks It Took to Make It Fast

Getting a self-hosted AI agent to hold a real-time voice conversation over Discord — on our own GPUs, with nothing leaving the network — turned into a two-week latency hunt through cold GPU handles,…

Infrastructure · Networking

The Felt Experience of a Well-Tuned Network

Most networks are utilities. A few are tools. The difference is whether you can feel the friction every time information passes through. You spend eight hours a day inside your network. More, if you…

Agents & AI

Agentic Web Scraping with Browser Automation

The Challenge I went into this project wanting to build an agentic agent that could scrape the web and populate web application forms with information. Traditionally, building a scraping agent is stopped by the…

Client work · Infrastructure

Fifteen Pipelines Into One

A full microservice estate rebuilt onto a single trunk-based pipeline, taking a release from most of a working day to thirteen minutes.

Client work · Infrastructure

Research Lab Infrastructure Modernization

Executive Summary Faced with 10‑year‑old Ivy Bridge servers, fragmented GPUs, and mounting energy bills, the lab launched a six‑month modernization from January to June 2025. The goals were to: Key results: Key metric Early 2025 June 2025 Δ Idle rack draw…

Next step

Want this kind of work on your team?

The same care goes into client work: written down, checked, and yours to keep.