I came into Unboxed after finishing my first year studying Computational Social Science, where there is a great emphasis on the synthesis of Computer science and Social science. Whilst I knew there was applicability of this synthesis in the real world, I was interested to see it in practice in an environment such as the one at Unboxed.
When I started, I was given a project brief on some ideas exploring the UK government petitions website, who work with Unboxed, and exploring metadata from available UK Govt petitions endpoints. Initially, it was slightly overwhelming as there were tons of possibilities and different avenues to explore, but with help from Ben, as well as Liam, Fede and Rhian, I was able to refine the project into a clear and achievable goal. This goal was to prototype a dashboard which displayed petition metadata and allowed users to filter by constituencies and categories, whilst providing insights into underused petition data.
I started by parsing through the petition endpoints, collecting the data I thought would bring greater insight into petition metadata. Once I had collected the data, I began creating endpoints on the FastAPI, such as;
We then had the idea to use AI to categorise petitions based on their title and description, but I had virtually no idea on how to build an AI classifier, so with help from Ben and several youtube videos, I had a crack at building my own AI classifier. This presented a large challenge, yet I managed to build something that at least resembled an AI classifier. Whilst it is most definitely not perfect by any means, including a substantial, and I mean substantial bias to assigning petitions to the category of “Economy, Work and Pensions”, it is something that I’m really glad I had the opportunity to explore and build. After I had built the AI classifier, I took its categorisations and connected it to the FastAPI.
Once I had created all the endpoints, I was able to do some work with Fede on design aspects for the frontend. This included doing something called “Crazy Eights”, where you fold a piece of paper into 8 squares (we only did it with 4 squares), and a timer is set for a minute for each square, and you have to draw a design for an aspect on the frontend that you think would be applicable. This really helped get ideas flowing and refine what I wanted to present on the dashboard. Following this design session, I started working on the frontend using React, which I had never used before, so I made sure to read its docs to prepare myself, then I got to work on the frontend. However, I was then presented with the challenge of time management, where I didn’t have enough time to complete the dashboard to the extent that I would have wished, so with a little (or large) amount of vibe coding - a term referring to building tools with the help of AI, I was able to come up with something at the end. This really taught me the importance of adaptability and adjusting to the needs of the project.
At the end of my internship, I had the opportunity to present my project to members of the petitions committee, which when I first heard about it seemed pretty daunting, but Liam and Ben helped me frame the narrative and work on my presentation. Yet here I am writing this blog, so I’m thinking it went pretty well.
Over the course of the past 4 weeks I have learned so much and couldn’t be more grateful to the team for all the help they’ve given me. It’s been great to have the opportunity to come in and work on something that I think is really cool and truly important. Along the way, I’ve developed some many new skills, such as design thinking, building AI models, using React and most importantly working in a professional environment. After my internship here, I plan mostly to enjoy the rest of the summer (not looking at too much code), before going back to university to start my 2nd year, where I look forward to building on everything I learnt.
