Artificial Intelligence Engineering (Human-Centered Artificial Intelligence), MS postgraduate personal statement example

Personal statement example

Most calls to the IT helpdesk where I work are about passwords. A tenancy officer rings because the system has locked her out again, and the fix takes thirty seconds. What takes longer is working out why the lockouts keep happening. Last spring I noticed that a cluster of calls came from staff using a new expenses tool that logged people out silently after ten minutes, so they returned to a form, lost their work and tried their password repeatedly. The software was functioning exactly as designed; it simply had not been designed around how people actually fill in expenses between site visits. I wrote this up for our systems manager, the timeout was lengthened and a warning was added, and the calls dropped noticeably. That small episode captures what draws me to human-centred AI: systems can be technically correct and still fail the people using them.

I studied Computer Science and graduated with a 2:1. The modules I enjoyed most were machine learning, where I first implemented logistic regression and a small convolutional network from scratch, and human-computer interaction, which taught me to treat usability testing as evidence rather than decoration. My final-year project brought the two together. I built a captioning tool that used an open-source speech recognition model to produce live subtitles for recorded lectures, then tested it with five hard-of-hearing students recruited through the university's disability support service. The word error rate I measured on clean audio was respectable, but the participants cared about different things: speaker changes going unmarked, technical terms being mangled, and captions arriving in bursts that were hard to read. I added a custom vocabulary list drawn from lecture slides and changed how text was segmented on screen. Neither change transformed the accuracy figures, yet participants rated the second version as considerably easier to follow. The project showed me that a model's benchmark score and its usefulness are related but not the same, and that I lacked the methods to evaluate the gap rigorously.

That is the gap I want postgraduate study to fill. I want a firmer grounding in the engineering of modern models, particularly how speech and language systems are trained and fine-tuned, alongside structured approaches to user research, explainability and fairness. Reading Don Norman's The Design of Everyday Things during my degree gave me vocabulary such as affordances and feedback, but applying those ideas to systems whose behaviour is probabilistic and sometimes opaque raises harder questions. How should an interface communicate that a caption is uncertain? When does showing confidence help users, and when does it simply add clutter? I would like to study these questions with proper experimental design rather than intuition.

Outside work, I volunteer fortnightly reading for a local talking newspaper, recording news and community notices for blind and partially sighted listeners. It has made me attentive to pacing and clarity, and listeners' feedback, passed on by the coordinator, is often blunt and practical: read the phone numbers twice. I also bake bread most weekends, which has taught me patience with processes that only reveal whether they worked several hours later, a temperament that has proved surprisingly useful when training models.

My helpdesk role has strengthened skills that a demanding course will need. I explain technical problems to people under time pressure, keep careful ticket records, write short scripts in Python and PowerShell to automate routine account tasks, and have learned to ask what someone was trying to do before asking what went wrong. I am now ready to combine that habit with deeper technical knowledge. I want to build AI systems that work well for the people who rely on them, and to be able to show, with evidence, that they do.