Projects
I have several active projects. REEL data is more complete and is a data management application. Woodendollars is a work management or organisational currency application to bring capitalist efficiency to feudalist organisations.
Woodendollars Link to heading
The Woodendollars concept has been brewing for about 15 years. If you work in an internal service provision team, as data and technology teams usually are, the prioritisation usually follows feudal or monarch styles. Outside of organisations more efficient and democratised approaches prevail. If a type of capitalism could be adopted within organisations efficiency could be improved by work being done where it adds most value. Service delivery teams could be chaned from cost sinks into revenue generators, all be it the revenue is wooden dollars.
Architecturally, Woodendollars is a web application with a relational back end with. I shan’t go into too much more detail here.
Unicast One Link to heading
Unicast is a user ventric communication platform to allow a sender to direct mesages to one address and the recipient can choose how they want to recieve that communication.
It came from a realisation I had when the NHS sent three letters for one subject. I didn’t need to revieve these as letters and email would have been both more convenient and significantly cheaper. If I could opt to recieve by email that would help both sides. There are some people who would still prefer to revieve physical latters or perhaps have the information on WhatsApp. Why not let them choose? So Unicast.one.
Demo bank Link to heading
MOP or Mortgage Origination Platform is a demo project built over a few days making extensive use of AI agents. I have worked in banking for several years and this was originally a demonstration to show what is possible with AI agents. I have since extended it to create an almost complete mortgage origination platform. It supports multiple tenants, multimple brands, multiple asset classes. It has a configurable, inherited lending policy engine which gives brokers clear, early feedback on applications. Workflow is supported by AI to analyse entered information against bank statements, payslips and credit bureaus.
It was built using specs with requirements and acceptance criteria rather than vibe coded. I used various frontier models in development as different models have their merits for different parts of the process. Architecturally it has a thin front end which proxies to a RESTful API with a relational data store. Deployment is automated with Github actions allowing changes to be tested and shipped to production in minutes. It also looks nice!
REEL data Link to heading
I originally wrote REEL data for a retail client where we were moving many tables from source to destination. Existing ETL tooling required a great deal of re-keying and was fragile. REEL allows for bulk loading of data pipelines, it supports moving pipelines between environments and it has a clear web based user interface. It was written in C# as the client was a Microsoft shop.
Recently I have migrated from C# to Go. Primarily so that REEL workers can run on any popular platform and so not require an additional runtime. It now also supports relaying via the central server so that workers don’t have to be on the same network segment as both source and destination data stores.
Architecturally, REEL data is made up of a central API with web management (hosted on GCP) then REEL workers are deployed close to the data source or data sink.