Dan Hill · Compliance Systems & Data Specialist
I work in food manufacturing compliance, turning complex requirements and manual processes into structured, reliable and auditable systems using data, workflow automation and better process design.
I approach technology as a practical tool for solving operational problems rather than as an objective in itself.
Outside work · self-directed projects, counted 29 Sep 2026
01 · Professional
Food Manufacturing · Compliance · Data · Automation · Process Improvement · Auditability
I am a food manufacturing compliance professional specialising in the intersection between compliance, data and operational systems.
My work focuses on turning complex requirements and manual processes into structured, reliable and auditable systems. I combine practical knowledge of food manufacturing compliance with strong capability in data management, workflow automation, reporting and digital process improvement.
02 · Outside work
Everything from here down is self-directed project work, done in my own time. It isn't employment. Projects & Development explains why I do it.
A Node server that steps a whole town forward every few seconds and catches up after an outage. A Discord bot that opens a channel for each new investigation. Render services for voice, video and podcasts. This site, with its own analytics backend.
Evidence
I use n8n as the conductor. The workflows I trust most are the dull ones: they check a lock before starting, log every outcome, refuse to act on bad data, and leave a trail I can read the next morning.
Evidence
A content pipeline went quiet for weeks. It wasn't broken. An approval email timed out and a switch with no fallback branch dropped the item without raising an error. Most of my fixes start by reading the real execution data rather than guessing.
Evidence
Gemini writes storylines and draws portraits and Polaroids for Arkham each month, and the server validates all of it before it goes live. A cloned voice narrates a podcast built from the town's own records. Posters are generated only once per event and then reused.
Evidence
Arkham exports a daily chronicle as a CSV. The café workflow pulls Meta post metrics back into a sheet every four days. Pi Ops keeps a resource diary so heavy jobs are scheduled from evidence. This site reports its own traffic every week.
Evidence
03 · Projects
Self-directed projects. Every number here comes from the running system, counted in September 2026. Nothing is a mock-up.
A persistent 1926 New England town that simulates itself in real time. You can watch it, but you can't interfere.
A board-game café's events, posters, captions and posting schedule, handled by a 178-node n8n workflow and three guarded sub-workflows.
Idea → script → human approval → images → narration → render → publish → anthology, as n8n pipelines running on the Pi.
14 services, 38 active workflows and one 8 GB computer. Resource-aware scheduling so nothing heavy starts just because a clock fired.
Voice cloning, animated video rendering, vertical Shorts and podcast rendering, all as small HTTP services on the Pi.
Websites, contact backends, email-triggered site updates with an audit trail, inbox triage and mail merges.
A portfolio with its own privacy-conscious analytics backend, admin panel, n8n reporting and Google Sheets data feed.
Where I've used it
04 · How I build
The same nine moves show up in every project. Pick a step to see where it happened.
01 / 09
Say what actually goes wrong, and for whom. 'Social media is hard' doesn't count. 'Past events keep getting promoted, and posters cost money every time' does.
For real: RTD: expired events were still being advertised. Roll The Dice Café: event and social automation
02 / 09
Where does the data live, who edits it, what decides, what acts. Usually there is one sheet or table that humans should keep editing, and everything else can be derived from it.
For real: The Event Index sheet stays the only thing a person touches. Roll The Dice Café: event and social automation
03 / 09
Anything that happens on a schedule, has a clear rule, or is copied from one place to another is a job for a machine.
For real: Date roll-forward, calendar sync and posting slots. Roll The Dice Café: event and social automation
04 / 09
Use AI where a person would otherwise have to be creative, and use deterministic code where a person would otherwise have to be careful.
For real: Gemini writes the caption. Code decides whether a post happens at all. Roll The Dice Café: event and social automation
05 / 09
Get something running end to end, however ugly it is. A working prototype shows me the problem better than a plan does.
For real: Arkham started as a single HTML file that ran in the browser. Arkham: A Living Survey
06 / 09
Webhooks, APIs, a shared table, a lock. The joins between systems are where most failures happen, so they get the most care.
For real: Arkham → relay → n8n → Discord bot, with each part able to fail without taking the others down. Arkham: A Living Survey
07 / 09
Log every outcome, including 'decided not to act'. When a run is silent I can't tell whether it did its job or broke without saying so.
For real: The Pi Ops resource diary and queue ledger. Pi Ops: a self-hosted platform on one Raspberry Pi
08 / 09
Check the real execution data, not the dashboard's summary. Before-and-after snapshots can hide what actually happened.
For real: A TFCD timeout showed up as a successful run. Only the node-level data showed the truth. AI content pipelines: The Ninth Volume, The Ninth Archive, TFCD
09 / 09
Fix the class of bug, not just the one instance, and port the fix to every place that shares the pattern.
For real: A TTS timeout guard fixed in Ninth Archive was ported to Ninth Volume the same day. AI content pipelines: The Ninth Volume, The Ninth Archive, TFCD
05 · AI
I don't use AI for the sake of saying I use it. It's a multiplier: it lets one person run what used to need a small team. The question I ask is what I can build with it, and what has to stay outside it.
Talking a design through, auditing a system, finding the one line that broke everything. Claude Code does a lot of the investigative legwork on the Pi.
Seen in: Credential audit, crash investigation, Pi Ops audit
AI-assisted engineering with me in charge: I set the constraints, review the output and insist on tests. Arkham's server, the Discord bot and this site were built that way.
Seen in: Arkham, Discord bot, this site
AI steps sit inside workflows with validation around them. The server checks a Gemini pack before anything goes live, and a failed image is skipped, not fatal.
Seen in: Arkham monthly update, RTD captions, content pipelines
Portraits, Polaroids, posters, scene art, narration in my own cloned voice, animated video, podcast episodes and clips.
Seen in: Arkham, RTD posters, Ninth Volume, podcast
Post metrics fed back into prompts, chronicle exports, resource diaries. The TFCD channel teaches compliance teams to use Copilot for audit evidence.
Seen in: RTD metrics loop, Arkham chronicle, TFCD
Self-hosted, including the AI services where that's practical: TTS and voice cloning run locally, and only generation calls go out to an API.
Seen in: Pi Ops, media services
06 · Architecture
One Raspberry Pi, one orchestrator, and a set of services that each do one thing. Click any part to see what it does and where I use it.
People
Edge
Applications
Orchestration
AI & media
Data & reporting
07 · CV & contact
I'm most useful where there's a real operational problem and a willingness to try something. I'm happy to talk about roles, projects, or building something for your business.
Curriculum vitae
Compliance Systems & Data Specialist: food manufacturing compliance, data, systems and process improvement.