Fjordvik
Aerial view of a calm Norwegian fjord at first light, with six round salmon pens in two rows of three and a feed barge

Concept demo · AI for salmon farming

Ask your farm.
Every number checked.

One assistant for the pens, the alarms, the harvest and the customers. It answers from the data, and code checks every number before you see it.

  • M4 lice 0.46 per fish, 92% of the limitChecked
  • M2 fed 16.8% above appetiteChecked
  • M5 oxygen down to 5.4 mg/LChecked
Pens on the site
6
Customers in 5 markets
25
Days of farm data
28
Screens to test
9

What it does

Not a mock-up. You can break it.

Screenshot of the dashboard with harvest, revenue, price and pen statistics

Production and sales in one dashboard

Harvest per month, price, revenue per market and how each pen is doing. Filter by market and every chart follows.

Open the dashboard
Screenshot of the world map with customers and a customer panel

See where the fish is sold

A world map of 25 invented customers, a customer database you can search and export, and a harvest planner. The AI weighs contracts, on-time delivery and claims, then proposes who gets the volume.

Open the map
Screenshot of the incident simulator with switches for a lice outbreak, oxygen drop, mortality spike and camera failure

Switch on an incident and watch it ripple

A lice outbreak in one pen, an oxygen drop, a mortality spike. Status, alarms, charts and the briefing all change. Then ask the AI what to do, approve its proposals and follow them on the task board.

Try an incident

How it is checked

The AI is not trusted. Its numbers are checked.

  1. It answers from the data only

    Pens, daily readings, alarms, customers and the harvest plan. A question about anything else gets a one-line refusal.

  2. It must name every number

    Each figure comes with its pen or customer, its measure and its value, so there is something to look up.

  3. Code looks each one up

    The server compares the figure with the data and stamps it. A wrong number shows up in red, a missing reading is called missing.

Checked M4 lice.latest = 0.46 Wrong M2 lice.latest = 0.31, data says 0.24

About

Built by someone who fillets salmon

I am Peter Heijnen. 30 years in fish. I founded Heijvis Visverwerking in the Netherlands, where we have always filleted a lot of salmon and still do, every day. Next to that I build AI agents and automations on Claude, n8n, Supabase and Cloudflare.

The same setup works for other fish companies, and for any business with production and sales data.

Daily filleting for seafood wholesalers in Urk and Spakenburg

Try it for five minutes

Nine screens, real AI answers, no login. Start with the dashboard or ask the site a question.

Open the demo

My CV · peter@lazylizardai.com