Eloqute have novel concepts – both pedagogical and technical – in the evaluation of spoken words and the teaching of pronunciation.The founder needed a demonstration of the pedagogical concept for potential investors.To this end I created a functioning web app that would allow the user to read from a text of their choosing and show ongoing evaluations. The prototype also had a number of other functions on the front end such as parsing the results,identifying the pathological phonemes.

Aegon are developing powerful data as service tools to re-purpose legacy data within the company, and to complement this they are providing tools to automate the building and deployment of applicationsthrough simple OpenAPI definitions to speed up and facilitate this process.I was placed within the API part of this team to help build and extend their core automation module, built in Node and deployed from an internal Artefactory repository.

Building an in browser transcription editor around a probabilistic lattice generated from an AI. Task was twofold:1) Build a surrounding application to load a transcription “lattice” and the associated media file from which the lattice had been generated.2) Build a library which would present the transcript to the user in an editable interface and update the transcript whenever the user made an amendment.This was a highly complex project allowing a user to interact freely with a state-of-the-art AI tool and a piece of audio simultaneously. Features included predictive word completion and audio seeking to the position when a word is clicked.Also built a custom test harness which would take a correct transcript and enter it character by character into the editor submitting words when the prediction was correct, thereby allowing the transcription engine developers to monitor the behaviour of their engine.The project involved highly dynamic custom event creation and the creation of tight fast code to keepwithin response times of speech-driven actions.Helped interview and board a permanent replacement, as was part of original agreement, and ahandover was effected with the new person.

Yubl was developing a new mobile social media concept for Android and iOS platforms. Thedeliverables included a web viewing platform which can show public instances which have been publicly shared to web users without an account or the native clients.To this end built a web application which runs from the shared API, using the node express and request modules to mediate between the browser and the API server, and using browserify to composite the viewer component which runs in the browser. This effectively makes it a micro service allowing uncredentialed clients access to publicly shared content.The main requirements for the front end of this is to create a clean component able viewer which canshow instances in a web page independently and without impacting on or by the content and design of the embedding page, and to create a framework which can handle any elements required to be handled by the public web view without breaking or halting when there are other elements present.The library was also used to render content for the moderating service interface which allows moderators to view and block content which contravenes the platform guidelines.The code was delivered fully conformant to the deployment pipeline, fully documented as specified and reviewed by the Technical Director and with a full set of unit tests for both microservice and client-side code using mocha, expect.js and super agent.

Macat had a requirement for a site to introduce their open public beta – there would be a strong marketing component, but it was also intended to capture pre-release signups to the beta both from invited student bodies and from interested visitors.Prefinery was used to capture the signup invitations and track levels of engagement and, uponcommitting to the full pre-beta signup, the user would be signed up to a mailchimp newsletter.These functions were performed using the public APIs of Mailchimp and Prefinery, called by ASP functions in the server-side application, with an AJAX API in the browser calling them in turn.There were two main objects in the browser: one for triggering visual marketing led animations andbehaviour according to user scroll and inpage navigation; the other an MVC-patterned application fortracking level of user commitment and status, communicating any changes back through the API, andmoving its view elements on accordingly. (There was also a small utility for managing social network http calls to allow the user to share a recommendation link which would be uniquely traceable to them, this was largely an utility).The application has been handed off and launched, and will have finished its work when the public beta launches with the users it has harvested.

Rewriting the Timeseries graphing module to account for asynchronicity and interrupt the graphing part with a set of options as to which datasets the graph would be drawn from. Also amend the graph drawing (d3) itself to reflect that there were in the new data format diverging degrees of confidence.

Created a Predictor from the sky sports betting feed which would allow users to place bets using a dummy currency. This was developed as an embedded in page application which could be loaded on any page consumng JSON from the sky feed processor via AJAX.

Developing further an in-page-application designed to deliver visualisations of data scraped and collected by the big data team from social media websites.The application was designed to give A and R, management and record companies an overview of their roster of artists with respect to their performance on social media sites such as youTube, Soundcloud, myspace facebook et al.The application would allow subscribers to follow their artists, and view their performance over time line graphs, bar charts, histograms etc.Main contribution has been to add track level data to the general artist level data, adding a “tracknavigator” for each artist which would allow subscribers to view charts of individual tracks rather thangeneral overview and overlay track data to compare performance over time.Also added a velocity chart and an acceleration chart per artist (a chart in the traditional sense of a top ten listing) showing their best performing tracks.

Relaunching Channel 4 Front page with a new application and design using data scraped from playout database and other sources to provide a web presence for every single programme regardless of whether they had a web content team.

Built the components which would allow journalists using the desktop editing application to drag and drop video content into the page and have it display in the published web page. I later generalised the component to handle any embedded content. Also was part of a small team which built the components an award winning front page redesign and relaunch.

Recruiting and heading a team to be more responsive to editorial needs and extending the contentproduction system to accommodate more varied types of editorial content.Scripts to produce web pages from MSSQL database for election data.Representing News New Media at several pan-BBC groups.