Artificial Intelligence Engineering (Computing Sciences), MS postgraduate personal statement example

Personal statement example

The self-checkout machines I service for a regional maintenance contractor flag 'unexpected item in bagging area' far more often than shoppers expect. Part of my job is recalibrating the weighing platforms, and after a year of visits I can usually predict which stores will call: those with heavy footfall, worn load cells and customers who rest a handbag on the shelf. The machines apply a fixed tolerance to every product. A bunch of bananas and a jar of coffee are judged by the same rule, although their weights vary in very different ways. I started keeping a simple spreadsheet of fault reports against store conditions, mostly to plan my rounds, and it made me want to understand how a system could learn tolerances from data rather than have them set once in a configuration file. That question, how learned models behave on cheap, imperfect hardware in public places, is why I am applying for postgraduate study in artificial intelligence engineering.

My undergraduate degree in electronic engineering gave me a solid grounding in signals, embedded systems and programming in C and Python. For my final-year project I built a keyword-spotting system that recognised ten spoken commands on a low-cost microcontroller. I trained a small convolutional network on Google's Speech Commands dataset, using MFCC features, then converted it with TensorFlow Lite for Microcontrollers and applied eight-bit quantisation so it would fit in the available memory. Accuracy on the test set fell only slightly after quantisation, but performance in my kitchen was noticeably worse than on the recorded clips, especially with the extractor fan running. Adding recorded background noise during training improved it, and writing up that gap between benchmark and real room was the most interesting part of my report. I was pleased that the finished device ran reliably on battery for a full day of testing, which had been my supervisor's main concern when I proposed it.

Since graduating I have been working through Deep Learning by Goodfellow, Bengio and Courville, slowly and with a notebook. The chapters on regularisation and optimisation have been the most useful, because they explained more formally why some of the tricks I used in my project worked. I am less confident with the probabilistic material in the later chapters, and I would value structured teaching in probability, statistical learning and the design of reliable machine learning systems, alongside the chance to build larger projects than a single microcontroller allows.

My job has taught me things that a degree did not. I work alone in shops during trading hours, so I explain faults to store managers who want their tills back quickly, and I have learned to give honest estimates rather than hopeful ones. I log every repair in a ticketing system, and the habit of writing a clear record that the next technician can follow has improved how I document code.

Outside work I ring bells at my local church most Thursday evenings and on Sundays. Change ringing is essentially working through permutations by memory and rhythm, and keeping my place in a method while eleven other people do the same is a good test of concentration. I also tutor two Year 11 students in maths each week. Explaining simultaneous equations to a fifteen-year-old in three different ways has made me more patient, and more aware of when I only half understand something myself.

I am applying for a master's degree because I want to move from repairing systems designed by others to designing them carefully myself. I am especially interested in efficient models for edge devices and in how to test whether a model will hold up outside the conditions it was trained in. I bring practical engineering habits, a completed embedded machine learning project and a working knowledge of what happens when technology meets ordinary, impatient users. I would like the rigorous foundation to do that work properly.