About

LanceTek is one person. That's the point.

No account managers, no hand-offs to a junior, no offshore team you never meet. The engineer you speak to on the first call is the one who writes the code, sets up the server and answers when it needs changing.

Who you'd be working with

I'm a software engineer based in the UK. I've spent my career building and running software — web applications, data pipelines, the servers underneath them — and I set up LanceTek to do that work directly for small businesses, where a well-made tool makes a visible difference to someone's week.

Small businesses tend to be badly served by software suppliers. Agencies are priced for larger clients; off-the-shelf products almost fit; and the person who "did the website" has moved on and nobody has the passwords. I'd like to be the opposite of that: a supplier who's easy to reach, explains things plainly and leaves you in control.

Being honest about one person

A one-person company has real limits, and you should weigh them. I can take on a small number of projects at once, so start dates are sometimes a few weeks out. I'm not a 24/7 support desk. And if I'm ill or on holiday, work pauses.

I plan around that. Everything runs on infrastructure in your name, the code lives in a repository you own, and every project ends with a runbook written so that another competent developer could pick it up. You're never locked in to me.

  • Smallest useful thing first

    Start with the step that saves the most time, prove it, then decide what's next.

  • Boring technology

    Well-understood tools with long track records, chosen so the system outlives fashions.

  • You own everything

    Code, servers, domains and accounts registered to your business, not to me.

  • Written, not just said

    Plans, quotes, changes and decisions go in writing, in plain English.

  • People decide

    Automations do the legwork. Decisions about money and customers stay with a person.

  • Say no early

    If an existing product does the job for less, I'll point you to it.

The stack

What client work is built with.

Chosen for long support lifetimes, healthy communities and how easily another developer could take over. The Journal explains many of these choices in more depth.

Applications

  • Python & DjangoAdmin screens and data-heavy tools
  • TypeScriptInteractive front ends, kept light
  • htmxResponsive pages without a heavy framework

Data

  • PostgreSQLThe default for almost everything
  • SQLiteFor small, single-site tools
  • RedisQueues and short-lived caches

Automation & AI

  • Hosted LLM APIsChosen per job on cost, quality and data terms
  • Local models via OllamaWhen data shouldn't leave the building
  • n8nFor simple glue between existing tools

Infrastructure

  • Debian LinuxStable, predictable servers
  • Docker ComposeRepeatable deployments
  • Caddy, restic, Uptime KumaHTTPS, backups and monitoring
The homelab

Where things get broken before they reach you.

I run a small homelab: a couple of quiet machines and a storage box. Every tool I recommend runs here first, often for months, so I learn how it fails, how it upgrades and how it restores — on my data, not yours.

homelab — logical layoutsimplified
Diagram of the LanceTek homelab A private VPN connects a laptop to the homelab. A small router and firewall sits in front of two hypervisor hosts. Host one runs a code forge, CI runners and staging copies of client-style apps. Host two runs monitoring, a local AI model server and a test bench for self-hosted tools. Both write to a storage server, which replicates encrypted backups to off-site object storage. Nothing from the homelab serves client production. Laptop anywhere, via VPN Firewall VLANs, WireGuard Host 1 · Proxmox build & staging Forgejogit + issues CI runnerstests on every push Staging appssame Compose files as production Host 2 · Proxmox tools on trial Monitoringmetrics + alerts Local LLMGPU, private tests Test benchself-hosted tools before I recommend them Storage server ZFS mirror, snapshots Off-site backup encrypted, UK region Nothing in the homelab serves client production. Client systems run on UK hosting or the client's own hardware.
# what "tested in the homelab" means, in practice 1 run it for at least a month on real (my own) data 2 upgrade it at least once, ideally across a major version 3 delete it and restore it from backup 4 write up what went wrong — that becomes a Journal entry
Work together

If that sounds like the kind of supplier you want, let's talk.

The first call is free and there's no obligation. You'll come away with a clearer view of the problem either way.