2023 list
Complete list of design briefs to be advertised to students for 2023 group design projects.
(text transcluded from individual project descriptions - click on project title to edit original page)
A Teaspoon of Video
Client: Michael Morehouse, MVES <michael@yawpitchroll.com>
Computer Science education expert Mark Guzdial suggests kids can learn effectively from Task-Specific Programming (TSP, or "teaspoon") languages that can be learned in a few minutes, teaching one idea from theoretical CS (the medicine?) in the context of a motivational and interesting automation task (like a teaspoon of sugar to help it go down). Your job is to create one of these educational languages, that teaches an important computing concept in a form relevant to processing social media video.
AI Hype Monitor
Tomasz Hollanek, LCFI <th536@cam.ac.uk>
Researchers critical of the persistent 'AI hype' – driven by flashy headlines and corporate communications – have pointed to the harmful effects that misrepresentations of AI in the media might have on public perceptions of the technology. Your task is to create an automated monitor of AI hype content, based on a predetermined classification system of particularly problematic metaphors and comparisons common in news articles about AI that circulate on social media. Ideally, the AI Hype Monitor could be used as an add-on filter for chosen social media applications, flagging up problematic content and pointing users to sources that represent current AI capabilities more accurately and responsibly.
Automatic Entrepreneur
Client: Daniel Organisciak, Luminance <daniel.organisciak@luminance.com>
The skill of a business entrepreneur is turning a few key terms, and some names of people and places, together with some financial figures, into a compelling narrative. Until now, it has been time consuming to manually extract relevant names and figures from reports filed at Companies House. It requires hard thought and creativity to write about the business opportunity. Your project has two parts. The first is automatic extraction of pertinent information from company filings, using methods such as syntactic parsers or named entity recognition. The second part is to use a combination of visual design and generative language models to create web pages that pitch the business to new investors, or perhaps provide the template for a competing start-up.
This project team has used public datasets via the API from Alpha Vantage: https://www.alphavantage.co
Autonomous Trucking
Client: Arlen Anderson, Ab Initio Software <arley@abinitio.com>
Autonomous vehicles are likely a wave of the future. A lot of effort is going into automating the control of individual vehicles to provide a self-driving version of our current experience. An alternative is centralized control of fleets of vehicles to optimize large-scale traffic flow. This is a simpler problem which might be very appealing for transport of goods by truck. The brief is to build a simulation environment to demonstrate centralized control of a fleet of trucks. The simulation should consist of three elements: a) a traffic optimization algorithm which achieves maximal transport of goods, given boundary conditions for trucks joining and leaving the flow of traffic, b) a visualization of the traffic flow, c) control of the traffic variables, including highway pattern (on-ramps, off-ramps), rates of trucks entering and leaving the traffic pattern, basic dynamic properties of trucks (e.g. maximum rate of acceleration/deceleration).
Don’t Stop the Music
Client: Al Wood, MQA <Al@mqa.co.uk>
Audio mixing desks are a mystery to many, as you might know if you’ve ever had to diagnose the silence when your friend’s band goes onstage. Desks have only got more complicated as they integrate USB soundcards for capture and streaming. There’s an opportunity to make life a lot easier for the amateur stagehand, with a phone app to communicate over Bluetooth with a standalone computer (say a Raspberry Pi) that can be connected to the USB port of any low-end mixing desk like the Yamaha MW/12CX. The app should listen to the sound output, analyse the signal, and send either audio test data or user instructions to optimise all the settings.
Electric Motor Optimisation
Client: Daniel Bates, Monumo: <daniel.bates@monumo.com>
Electric motors are all around us and together consume about half of the world's electricity. Even small improvements can therefore have large global impacts. Your task is to streamline the early experimental process of motor design, allowing a domain expert to prototype a good design more quickly and/or produce a better design before they reach their deadline. This will involve building an intuitive interface for designing motors, learning about the underlying computational models, and analysing and displaying the results in a way that highlights important features.
Email Diplomacy
Client: Angus Allen, Volemic <angus.allen@volemic.com>
It is easy to make a mistake when sending an email and the consequences range from personal embarrassment to professional misconduct. In order to prevent these errors, it is likely to be helpful to understand the relations between different people associated with an email. At its simplest, this could involve identifying and classifying the sender, each recipient and any other people mentioned in the body of an email. It might also be useful to provide a confidence score for each relation, to recognise where a recipient is a member of a group of people such as an organisation, or classify the tone of the language used to ensure it is appropriate.
Fossil or Future?
Version for 2023:
Client: Matthew Postgate, Informeta Ltd <matthew.postgate@infometa.com>
Humanity needs to respond to climate change, but there is a lack of science-based policy for sustainable action. Legislation is not consistent, and corporate actors continue to lobby against climate legislation even while promoting environmental strategy in other areas. These factors make it hard for politicians to make essential choices between fossil fuel strategies that worked in the past, and the unavoidable future. The goal of this project is to develop a science-based decision tool, for use when drafting and reviewing policy proposals, that both analyses the language used and calculates realistic projects along the lines of David MacKay’s classic book Sustainable Energy Without the Hot Air.
Older version discussed for 2022:
During the COP26 conference, it became clear that many politicians around the world do not necessarily follow the party lines of the 20th century left versus right, but are dividing themselves between those who mainly plan for the future, and those that focus on protecting the past. It should be possible to use natural language processing to help the electorate decide between these alternatives, by identifying the real allegiance of public officials to classify them as either fossil or future. Simple bag-of-words classifiers won’t be enough, because anybody could mislead readers by dropping in random words. It’s likely that you’ll need to use some kind of deep learning, and also plan for the possibility of inadvertent bias or even malicious sabotage when collecting the training set.
Genomics for the COVID Endemic
Client: Dan Bolser Geromics <dan@geromics.co.uk>
Open genomics provides an opportunity for members of the public to engage directly with genomic health research including Genome-Wide Association Studies (GWAS). Now that the emergency response phase of the COVID pandemic is settling into management of endemic infections and their consequences, there are many opportunities for members of the public to contribute to such studies. The goal of this project is to develop an interactive genome report that keeps members of the public informed, and even allows them to participate as citizen scientists in ongoing investigations of the many consequences of COVID.
iMoji
Client: Patrick Wollner <patrickwollner@gmail.com>
It is useful to include emojis in your messages as a quick indicator of emotional state, but why should you have to call up a special keyboard, or scroll through many alternatives, when your emotional state could be read off your face? The goal of this project is to augment the on-screen keyboard by using the front facing camera (and/or depth data provided by the device) to just read off the emotional state and put the "right" emoji in. It could also be useful to use other sensors of the device to better assess the emotional and physical state of the user in selecting the fitting emoji.
Just Maps
Client: Clara Todd <clarabara@gmail.com>
Cambridge has been named the most unequal city in the UK, with homelessness, poverty, and unaffordable housing exacerbated by unequal access to land. The goal of this project is to produce an interactive visualisation that combines data from the Cambridge Land Justice campaign with census data, pollution data, the Green Space Index, and crowd-sourced reports on public access. This project provides an opportunity to interact with local groups advancing spatial and social justice in the city, while applying data science to real problems.
Keeping Key-Workers
Client: Konrad Maliszewski, Atheon <konrad.maliszewski@atheon.co.uk>
The design strategy of “gamification” uses familiar features of videogames like high scores, personal bests or daily streaks to improve incentives and enjoyment of everyday life. In the retail logistics sector, especially key-worker areas like food retail, there are many data feeds and performance measures available. Some retail sites do provide game-like feedback to staff and customers, but there are many opportunities to make this enjoyable rather than an oppressive obligation. Your client is a leader in retail data science, and you’ll have an opportunity to work with real data from real stores to see how key-worker’s lives might be improved.
Likeness Trainer
Client: Zafiirah Hosenie, Microsoft Mixed Reality and AI Labs
The human visual system is highly evolved to recognise faces even from a few lines. AI systems are improving, but we need more efficient tools to communicate which lines are meaningful to humans. The goal of this project is to create a tool for drawing the most essential lines onto photographed faces, as a starting point for psychological research into this fascinating human skill.
ModularSynth.io
Client: George Welch, IMC <george.welch@imc.com>
Sophisticated digital music composition tools like the Sonic Pi language rely on an internal architecture of samples, waveforms and filters. In the popular SuperCollider system, a new synthesiser is defined by software-wiring together these "UGens”. Your task is to create a SuperCollider client that looks like a retro-style modular synthesiser or guitar pedal board, where connecting literal wires between pictures of hardware modules on the screen will construct an exact digital equivalent within the SuperCollider server. A live audio input would give you a universal guitar pedal, sample mixing makes you a DJ/producer, or if bleeps and whooshes are your thing, you can impress your Grandpa by channelling Brian Eno in the glory days of Roxy Music.
NFTs for Digital CVs
Client: David Sharp, International Workplace <dgs48@cam.ac.uk>
When people move jobs they take the knowledge they’ve gained with them, but their learner records often stay with their employer. This is not only impractical, but ethically questionable. The advent of Web 3.0 technologies allows people to take ownership of their learner records and control who they share their data with. Utilising SoulBound Tokens (SBTs) your task is to build a model that captures a user’s learning engagement data, certificates etc. via a non-transferable NFT smart contract (known as a ‘Soul’), based on Ethereum. Your smart contract code will likely be written in Remix IDE and Sol/Python language with “mint” and constructor rules. Another challenge is to think about a data structure standard and how data exchange and editing will be done. There are two use cases: 1) when someone wants to share their Soul with their current employer; 2) when someone wants to add or update data on their Soul. The main priority is to make this process safe and secure for the user.
Physical Computing for Beginners
Client: Tracy Gardner, Raspberry Pi Foundation <tracy@raspberrypi.org>
Children and other beginners can understand the computational concepts involved in physical computing (controlling electronics components) before they are ready for the technical APIs provided to control hardware devices. Your task is to design and prototype a visual language or interface for programming the Raspberry Pi Pico microcontroller that can be used by beginners. The language should use concepts that are close to the kinds of projects that beginners would want to create. Users might drag and drop components such as buttons and LEDs and then click on them to configure how they should behave.
Rational Trading
Client: Patrick Wollner <patrickwollner@gmail.com>
The valuation of investments according to their expected future value is a key part of our economy. Unfortunately, sometimes investment decisions are made as joke or in the absence of a justified reason, as in recent scandals over GameStop and Dogecoin. The goal of this project is to create an alternative investment platform in which every trade is securely associated with a reason for the valuation, including links to public sources. Other investors should be able to make judgements based on individual trends of the asset, overall markets trends and the aggregated assessments of the reasons given, based on natural-language processing methods, to help assess how serious the opportunity really is (i.e. if associated risk of investment is justified by the reasons given).
Real Life Arabic
Client: Saussan Khalil, Kalamna <smk58@cam.ac.uk>
Mother tongue literacy is a major problem for young people across North Africa and the Middle East. The classical Arabic they are taught is like an Italian learning Latin. Apps like Duolingo do not help learners connect with the everyday language. They need a language learning app that enriches their lives with “kalamna” - the Arabic for “our words”. Your task is to design an app with motivating game functionality that connects to the Arabic needed in real life.
Responsible AI Copilot
Marios Constantinides, Nokia Bell Labs <marios.constantinides@nokia-bell-labs.com>
Many AI-driven applications turn out to have built-in biases, or problems of trust and transparency. In principle, developers could be warned of such problems as they write code, adding additional code to address the problems, recommending a specific debiasing algorithm, or adding inline comments or sticky notes that warn of need for future action. Your task is to provide such facilities in a modified version of Jupyter notebooks, perhaps using generative language models such as GPT-3 or OpenAI Codex to generate the relevant code and text output.
Responsible Construction
Client: Akvile Valentukonyte at Qualis Flow - <Akvile.Valentukonyte@qualisflow.com>
Collecting accurate data about materials entering and leaving a construction site is essential to ensure minimal waste and climate-responsible behaviour. However, most of the available information arrives on paper documents or pdf receipts. Optical character recognition helps identify what is written on the tickets, but the challenge is reliably identifying actionable information like arrival times, certifications, and especially key data that has *not* been captured. Your task is to create a phone app that can instantly capture and advise drivers and gate staff, while populating a dashboard that can be presented to local residents on a public display.
Robot Farm Monitor
Client: Kyle-James Keen, Autopickr <kyle@autopickr.com>
Much of agricultural work is difficult and dangerous, leading to a labour shortage across the sector and an increase in robotics and automation on farms. Growers today commonly employ cutting-edge technologies including temperature and moisture sensors, robots, and GPS technology to perform the tasks that humans don’t want to do. Your task is to design an app-based system that reports sensor data and health and safety information from our asparagus picking robot, over both The Things and Helium IoT networks.
Signs of our Times
Client: Katherine Powlesland, Digital Levi Project <klp47@cam.ac.uk>
In times of political turmoil, media commentary is transient, with more profound understanding coming only decades later. In future years, we will find that someone really knew what was happening today, in the same way Primo Levi, for example, captured the true tragedies of the Second World War in his masterpiece If This is a Man. Levi invites us to vividly participate in his text so that the same thing is never allowed to happen again. What future tragedies might be avoided if we could bring scholarly analysis of important works like these into dialogue with current concerns and debates, especially as expressed online? Your goal is to build a prototype for an innovative digital edition of If This is a Man, that inspires thoughtful reflection through immersion and elegance far beyond the trivial claims made for the Metaverse.
Speak in the Country
Client: Alex Raymond, Speechmatics <alexra@speechmatics.com>
People who live in rural areas of low-income countries find it difficult to access online services. They may have low literacy, and often rely on feature phones rather than smartphones for their communication needs. The goal of this project is to help people in that situation access medical advice via the latest speech technology from Cambridge company Speechmatics, in a system that will take voice messages in a local language, transcribe and match their needs to relevant information resources, and return recorded advice or monitor disease outbreaks. Health workers at local clinics should have a dashboard to monitor issues in their region, and also to flag conversations where a human expert needs to intervene.
