Dissertation, Direction, and Career

Looking Forward  ·  MSc Computer Science (Conversion)  ·  Orville Fernandes

The Dissertation

The project I am building for my dissertation is an AI-assisted shoplifting detection and evidence management system. The core idea is a retail loss prevention tool designed specifically for small convenience stores and corner shops: a system that monitors CCTV feeds in real time, identifies suspicious behaviour, alerts staff, and automatically stores and labels evidence for later review or reporting.

The problem space came from COM7119, where the coursework involved designing a retail loss prevention system at the scale of large retail chains. I found the concept interesting, but the original scope had a problem: building something credible at that scale requires insider knowledge of enterprise retail operations that simply is not accessible to a student working independently. Rather than produce something superficial to fit an unrealistic brief, I pulled the scope back. A corner shop is a more honest target. It is also, I would argue, a better one. The problem is real and immediate for a single shop owner trying to manage a store with one member of staff in the evenings. The value of a system like this is tangible and specific, not a theoretical exercise dressed up as industry relevance.

The other requirement I had for my dissertation was that it had to involve actually working with AI models – training them, running them, evaluating them. Not reading about AI, not describing AI architecturally, but building something that uses it. The shoplifting detection problem requires exactly that: person detection, behaviour classification, cross-camera tracking, and a continuous improvement pipeline that feeds validated events back into the model. That combination of a clear real-world problem and genuine technical depth is what made me settle on this project.

What I Am Building and Why It Matters

The system has three layers: a CCTV input layer feeding into an AI detection layer that handles person detection, tracking, and behaviour classification, which then pushes alerts and evidence to an application layer where staff can validate events, review footage, and access a structured evidence repository. The continuous improvement pipeline loops validated events back to the model, so the system gets more accurate the more it is used.

I am realistic about what a prototype built in a few months can achieve. The accuracy of the model will have limits – that is expected and honest. But I fully expect to deliver a working system overall, and I think the design is sound enough that with more time and data it could become something truly useful. Whether that means developing it further as a product, or whether it becomes a starting point for a conversation with a company working in this space, depends on where the results land. Either way, it is a piece of work I want to be able to show people.

What Comes After

My background is in cloud engineering – Docker, Kubernetes, CI/CD pipelines, infrastructure. What I want from a career is not just to be competent at a fixed set of tools. I want to keep learning. One of the things I genuinely enjoy is picking up a new technology, understanding it properly, and finding a meaningful way to apply it. I do not want to be the person who has done the same thing for five years and calls it experience. I want the kind of role where learning is part of the job, not something you do in your own time if you can manage it.

The other thing that matters to me is what the work is for. I want to be somewhere that is trying to make things better – for its customers, its users, or the world in some wider sense. Not as a marketing line, but as the actual reason people show up. A team that cares about the quality of what it builds, and cares about who it is building it for.

My interest in AI has grown considerably over the course of this programme – more than I expected when I started. Working with models, understanding how they are trained and evaluated, thinking about where they add value versus where they are a solution looking for a problem: all of that has become something I want to continue developing. The dissertation is directly in that space, and I think it positions me well for roles where AI is part of the work rather than just a talking point.

The dissertation is part of how I get there. A working AI system with a clear use case and documented methodology is a stronger portfolio piece than most coursework assignments, and I intend to make it as strong as I can. But beyond the portfolio, I think the act of designing and building a full system independently – scoping it, making the hard decisions about what to cut, dealing with what does not work – is its own kind of preparation. It is the kind of work that tells you something real about how you think and what you can do.

A Note on Confidence

Something has shifted over the course of this year that I did not entirely expect. Engaging with the work – the coursework, the conversations with lecturers and peers, the things I have built – has made me feel more confident about my own ability than I did when I started. Not unconditionally confident. I still have moments of impostor syndrome, moments where I wonder whether I belong in the rooms I am trying to get into. But there is less of it than there was. What I can say, going into the final stage of this degree, is that I feel competent. I think I have what it takes.