Skip to content

Work

Six projects, and the part AI really played in each.

No client is named here. Some of these projects were built before AI assistants existed, some are now maintained with them, and one was designed with AI from the first line of the specification.

01 / 06

Ship construction site visits and weight tracking, on a tablet

Three applications, one workflow: prepare on a PC, record on a tablet at the yard, finish the report without retyping anything.

Role of AI :AI from the specification onwards

140 hfrom specification to the first version in production

Sector
Naval architecture
Technologies
AndroidiPadWPFWindows 11

Context

A tablet application for following the construction of ships and the evolution of their weight estimate, for a company of engineers and naval architects.

Challenge

When the engineers and naval architects visited the yards building the vessels they had designed, they took a laptop for notes and a phone for photos. It was impractical on site, it was hard to be sure everything had been recorded, and each visit was followed by a long write-up.

What we built

A three-part solution: an Android application optimised for tablets, an iPad application, and a WPF application for Windows 11. Preparation and follow-up are done on the PC, where it is faster, but can also be done on either tablet. The work during the visit itself is done on the tablet only.

Role of AI

This is our first project in which AI was used from the drafting of the specification itself. Only 140 hours of work were needed to put its first version into production.

02 / 06

Predicting CO2 and fuel savings from wind-assisted propulsion

A solver that tells a shipowner what wind-assisted propulsion systems really save on its merchant ships, in near real time.

Role of AI :Now developed with AI

Sector
Shipping
Technologies
PythonMicrosoft FabricDatabricksGitHub

Context

A tool for a shipowner that predicts the reduction in CO2 emissions and the fuel saved after installing different types of wind-assisted propulsion systems (WAPS) on merchant ships.

Challenge

Shipping relies on vessels that mostly run on fossil fuels, with very high CO2 emissions. This client has set itself the goal of zero CO2 emissions before 2050 and is trialling a great many technologies to get there. It needed a tool to check the real benefit of the technologies fitted to its ships.

What we built

Python code built around a solver, which gathers its inputs from CSV and Parquet files and from Microsoft Fabric lakehouses. It produces analysis reports and charts, and exists in a special Microsoft Fabric notebook version able to run the CO2 and fuel-saving calculations in near real time.

Role of AI

The project began without AI. It is now fully connected to Claude Code and Codex, and to GitHub, since several companies are involved in its development. The real data and the credentials are confidential and are meticulously kept out of reach of Claude Code and Codex.

03 / 06

A bespoke document management system, maintained since 2010

Extended every year since 2010, and now receiving features that had waited for years because they cost too much.

Role of AI :Now developed with AI

Sector
Engineering · Superyachts
Technologies
WinFormsC# / .NETCloud storage

Context

A custom document management system (DMS) for engineering teams in a large British group and for a company specialising in the naval architecture of superyachts.

Challenge

The software is a WinForms application whose first version was written in 2010. It has been maintained continuously, with new features added every year. It now manages files stored in the cloud, and makes up for the availability problems that come with that kind of storage.

What we built

A Windows desktop application for managing engineering documents that are now mostly cloud-based. It has been extended year after year rather than replaced.

Role of AI

Developed without AI, the source code is now connected to Claude Code. Increasingly complex features can be added much faster and at a reasonable cost. Some of them had been identified as needs for several years but would have been too expensive to develop without AI. They had been set aside, until we implemented them this year.

04 / 06

An Excel add-in to track bills of materials and their cost

When raw material prices move with world events, production costs and quotes have to follow. This add-in keeps track.

Role of AI :Maintained with AI

Sector
Composites manufacturing
Technologies
Excel add-inCompany data sources

Context

An Excel add-in that tracks the evolution of bills of materials and their cost, for a company manufacturing composite parts.

Challenge

This industrial client faces variations in the cost and availability of certain raw materials and products, depending on the geopolitical situation of the moment: Covid, wars, climate disasters. These variations affect its production costs. To follow them over time, it needed a tailor-made tool.

What we built

An Excel add-in that is easy to install on the workstations of the people who follow production costs and quotes. It connects to several of the company’s data sources to retrieve the information it needs.

Role of AI

The project was coded without AI. It is now maintained with Claude Code, taking care never to give Claude Code any secret that would let it connect to the company’s data. The source code itself is not critical, so it is exposed to the AI, which allows new features to be added on a very tight schedule and a tightly controlled budget.

05 / 06

Production planning for a sail loft

A web application that shows where the loft floor will run short of space, early enough to keep every delivery date.

Role of AI :Built before AI

Sector
Sailmaking
Technologies
BlazorC#APIsAzure

Context

A web application to help plan the manufacture of sails. A Blazor and C# project, integrated into the company’s existing software through APIs.

Challenge

The sail loft needed a tool to plan its production better and to spot, as early as possible, the bottlenecks in the availability of the loft floor, so as to meet the contractual delivery date of every sail ordered.

What we built

A web application connected to the API of another of the company’s web applications and to a database. Different people in the company sign in to manage the schedule, or simply to look at the volume of orders in the near or more distant future. It is deployed on Azure.

Role of AI

This project was built without AI, which was not yet available. It was coded in three months. It would take a few weeks today.

06 / 06

Analysing very-high-frequency sensor data around a foil

Tens of millions of sensor readings, dozens of analyses, and automated reports to tell separation, ventilation and cavitation apart.

Role of AI :Built largely without AI

Sector
R&D · Racing yachts
Technologies
PythonAutomated reports

Context

A RAPID R&D project (the French scheme supporting dual-use innovation): analysing very-high-frequency sensor data to study the hydrodynamic flow around a foil, in collaboration with a French industrial SME and a very large French group.

Challenge

More and more racing yachts are fitted with foils that lift the hull above the surface of the water. These boats reach very high speeds and experience several kinds of hydrodynamic phenomena: separation, ventilation, cavitation. The aim of the project is to analyse sensor data and to try to isolate and identify these phenomena.

What we built

An analysis tool in Python that takes tens of millions of data points as input and runs dozens of different analyses, on the whole dataset or on targeted samples. It produces automated reports and charts of the results, so that they can easily be compared with one another.

Role of AI

This project was carried out largely without AI, apart from a few prompts to improve targeted components, without ever giving the overall context of the project or exposing the whole of the source code.

Clients we can name

These companies have agreed to be cited as clients. The projects above are not attributed to any of them.

A need, an idea, a spreadsheet that has outgrown itself?

Tell us about it in a few lines. We reply by email, and can follow up with a video call.