This was a 6 month project to help an ecommerce analytics company explore the quality of their scraped web data that drives their influential metrics.

I designed a prototype process to ingest and compare daily data against rolling averages and set acceptible boundaries for the quality of the data. These metrics could then be compared against a whole sales platform (Amazon for instance) for a specific retailer.

We worked together to understand the needs of the users, the sales platform maintainers and the data acquisition teams and employ a dashboard to demonstrate actionable insights to users.

I have developed several machine learning models for the same client that have been built around the property lending knowledge I gained at Brismo.

The mainstay of the models has been a metric that I designed to help differentiate loan applications based on key inputs. Implementing this metric has allowed for much easier classification of applications into types.

The models were initially deployed as APIs on a server hosted on a PaaS provider. This allowed for quick testing and rapid changes. We initially used the Azure suite of ML tools to train the models. We have not transitioned to an AWS offering and have re-tooled the models to be served through AWS lambda functions.

This is a project that I've been working on to take a product from idea stage through product launch.

Currently the product is in the minimum viable product stage and is being tested.

More information about the project can be found in the projects page.

This was a fun short project to collect data and create insights from proximity data. One of the issues about proximity data is that items can be close to each other but fall in different counties or be difficult to reach with existing transport links.

In this project I used uber's h3 to plot items across the UK and then used some bespoke algorithms to indicate the relative importance of the concentration of proximal features. I then plotted this using the mapbox tools.

Initially I used the Google reverse geocoding APIs to turn a location search into a lat/long value that could be plotted but then I transitioned this over to use the mapbox APIs to keep the solution code within one App.

Working for a small but influential investment company, I built a bespoke CRM tool that would allow them to capture information from recent meetings conducted. Their requirements meant that an off-the-shelf tool would take longer to implement and would be less efficient than one built specifically for their purposes.

Initially this was built in an excel document to work at the lowest common denominator. It allowed for communicating within their security requirements. I implemented communication with the database backend through VBA code.

This solution was good but the forms were not where they really needed to be, i.e. Outlook. I retooled the product to work as a VBA pluggin for outlook so that the forms were present in their email systems.

I progressed this product to be a fully fledged native desktop application by creating an ElectronJS application. While we didn't get the opportunity to role this out for the team, it was a fantastic opportunity to learn the power of Electron to build tools with a low footprint codebase.