Thomas McCallum

The objective was to allow students to monitor their progress throughout a range of projects towards fulfilling a set of learning outcomes.

The application was comprised of a WordPress plugin written in PHP 7. This design decision was chosen as it enabled us to utilize existing capabilities of WordPress that were not central to the core requirement. It had a separate MySQL database to store data about the student's work.

A scheduled cron task was created to regularly download the student commits from the repositories and store this data in the database. To provide meaningful visualizations to the student an R background process analyzes the student progress and maps this against an expected trend.

The tool used a Docker instance to allow for a clean environment for unit testing.

The challenges were around picking a framework and technologies that could be continued by staff after development ended. It also had to fit in with existing technologies that the university used, which constrained the design.

This was developed as an internal project and therefore is not available to the public.

This was a piece of work that undertaken after a high availability cluster kept failing loads over causing high priority incidents. The objective was to provide a way to allow those in charge of managing the load on the cluster to match loads in such a way that they would not cause a failover. Due to the highly secure nature of the environment, the tool was created using Excel VBA to create a suitable simulator.

At the heart of the simulator was a genetic algorithm that optimized the workloads. It consisted of a set of shell scripts that took live data for the performance of each application on the cluster (memory, CPU, and IO). From this data, features were extracted that were used to optimize what loads could work together best.

This tool was developed for a secure environment and therefore source code is not available.

Due to the highly secure nature of public sector estates, it is hard to get new software on to them. This led to large amounts of manual effort being required to do relatively simple tasks that today would be taken care of by products like Ansible and Chef.

I developed a comprehensive set of bash scripts for automating a range of maintenance activities on a large highly secure estate without the need for additional application installations. The script was able to take another script written for a single application of the task and then run this across a set of machines in parallel highlighting errors when applicable. It handled all the login and security access requirements. These scripts were also later adapted to also run using PowerShell.

This solution led to tasks that did take days to minutes. It also allowed the team to provide security reporting that otherwise would have taken weeks to get the information for.

These tools were developed for a secure environment and therefore source code is not available.

The challenge was to produce a dashboard that would allow a large property company to identify new land purchases for care homes.

The solution I developed was a web dashboard written in PostgreSQL and PHP (back end) and JavaScript (front end). It used R to do the analytics in an offline process.

The application processed data from the Office for National Statistics, NHS, OpenStreetMap, Land Registry, and other public data sources. Due to the changing nature of the data from these sources, the tools were flexible in how they could analyze the structure of the data and export information in a way that was in a consistent format that could be analyzed.

As an additional requirement, it was possible to generate brochures for existing properties from this information through a pipeline developed in VBA. The reason for using VBA in this instance was that the brochure inputs and outputs needed to be modified by non-technical individuals who were able to use Powerpoint and Excel.

This was an internal project contracted under NDA.

The challenge was to create an analytics platform and dashboard that could take in data from HFN and Lippa and query performance details about individual or groups of hedge funds.

The solution was a web-based tool that allowed the user to mix and match financial instruments with hedge fund performance trends. Having taken in data from Bloomberg, HFN, and Lippa. It corrected the data and made it accessible for benchmarking and clustering. A range of mathematical analyses was run on the data to allow for investment decisions to be made.

The solution required a number of new modules to be created for R in C/C++ to help analyze the quantify of data.

Technologies included: R, PHP, MySQL, C++ (for creating native R libraries)

This was an internal project contracted under NDA.