Seán Healy

My research focuses on problems of intersectionality in fair personalised information retrieval (IR) systems. Current fair solutions in machine learning and IR are often ill-suited to address more nuanced forms of bias, such as discrimination against subsections of larger protected groups. In my research, I develop and study novel algorithms and systems with a focus on protecting intersectional groups from biased treatment.

1. At EdgeTier I primary worked on fulltext search and indexing. In the process, I gained a deep understanding of PostgreSQL, along with Python search and ORM libraries.
2. I also contributed heavily towards the abstraction of EdgeTier's backend system in order to support a multi-tenanted architecture. This is a direction any SaaS company must take in order to speed up the onboarding of new clients.

1. I improved several data-intensive endpoints, including InSchool’s roll call and absence registration pages. These pages are used by teachers across the Norwegian public secondary school system (taking roll call with a tablet or computer). My enhancements, taking load times from multiple seconds to milliseconds in many cases, has at this stage saved Norwegian teachers days of buffering time at the beginning of class.
2. I was the bash and regex guy in the office, often tasked with writing scripts to modify huge portions of code, automate tedious tasks, or enforce a coding standard rule across different languages. I led the charge in transitioning date and time standards used within Visma’s InSchool project to ISO-8601. This involved huge collaboration across teams.

1. I worked on the airline pricing team, and wrote a web crawler to download, index, and cluster pricing rule documentation (tens of thousands of pages of legalese text). This aided other software engineers in retrieving important information, and recognising related yet vague terms such as ’Category 4 fare’ and ’Record 3’.
2. I pointed out issues with a deletion procedure on the Linux servers used by the pricing team. The Linux kernel keeps an unlisted copy of a deleted file until every program reading that file has closed it. For months, Datalex’s QA servers kept running out of disk space, despite deletion procedures in place to remove old files. My Linux knowledge got to the bottom of this unexplainable bug.

1. I improved Webio’s underlying REST API’s readability and maintainability, measured by a 40% reduction in lines of code, by introducing object orientation and static utility functions in areas with code repetition.
2. I increased sales to new clients by rapidly creating high-demand features. I successfully oversaw the creation, for example, of a smart reply feature, a versioning system for Webio’s core functionality (the bots creation page), and more mission critical features like the ability for human agents to easily intervene in a conversation handled by a poorly performing bot.