skip to content

Department of Computer Science and Technology

 

Lent term 2022 Design Briefs

Carbon Accounting

Client: Matthew Postgate, Informeta Ltd

Today’s personal bank accounts and banking apps all work by counting money, a simple technology that was created centuries ago as a social consensus to support trade. In the future, we are going to need a carbon economy that realistically prices everything in relation to the actual costs for the planet. Your task is to create a prototype of a new banking system, including personal savings and payment apps, where every transaction is based on the verified carbon cost of products and services. Consensus on carbon costs should allow for initial estimated quotes the first time a particular transaction is made, with better calibrated scientific evidence funded by arbitrage to increase eventual accuracy for integrity of the whole system.

Conservation Evidence Synthesis

Client: Thomas White, BioRISC 

Many published descriptions of conservation projects contain information about the actions that were taken, what they cost, and how effective the results were. However, it’s impossible for one person to find and digest the corresponding pieces of evidence when the relevant publications can be so varied in format and length. Your task is to use natural language processing methods to assemble the most important quantitative and qualitative data, and generate short and approachable texts communicating the essentials of each publication.

Creative Community

Client: Matthew Johnson, Frontier 

Children today struggle with mental health challenges, and it seems from research that the social dynamics of platforms like Instagram do not help. There are alternative approaches to building community, for example the way that TikTok encourages users to build on each other’s contributions with memes, challenges and reaction videos rather than just accumulate likes (see https://bit.ly/3DiOIyj ). Your task is to create a mobile app that develops community through creative quotation and remixing, with feedback mechanisms that emphasise quality rather than quantity. TikTok already exists, so this app will be focused on phone-based drawing, painting and image remixing, rather than video.

Empathetic Chatbot

Client: Lee Wilson, Centre for Policy Futures 

Chatbot dialog builders can be used to create systems that are very good at dispensing accurate information such as medical advice, following a configurable decision tree, but the output text needs to be standardised and impersonal. Language model-based text generators on the other hand, produce text that is grammatical and possibly entertaining but often factually wrong. Your task is to integrate the two into a demonstrator for the World Health Organisation that can be configured by public health clinicians to give accurate advice when needed on problems like Covid infection, and can also stimulate mental health with original creative responses - but only when the question from the patient makes it safe to do so!

Exhibition Inference

Client: Neta Spiro, Royal College of Music 

The Royal College of Music Museum has one of the richest collections of music-related objects in the UK and Europe, spanning over 500 years of musical activity. A new layout has just been created, and the museum needs to learn how visitors respond to it in order to refine the design in future. We suggest tracking (with permission) visitors who may use QR codes to access media or links from a specific exhibition case. Your task is to combine time-stamp records of QR access with dead reckoning from a phone's step counter, glimpses of GPS signal through the museum windows, prior knowledge of valid walking paths around the museum display, and any other sources of data you can find, to get the best estimate of how visitors travel around the museum, as well as which exhibits they pause at and in which order.

Flyathlon

Client: Adham Ashton-Butt, British Trust for Ornithology 

Major sporting events such as marathons and triathlons often raise funds and awareness for good causes including conservation. This project is an opportunity to design a new global sport where human challenges are measured against the challenges faced by wildlife such as migrating wild birds. You will use the Strava API to capture records of human performance, and monitoring data from bird migration (available from your client) to represent wildlife. You will need to define the rules for this new hybrid sport, including secure and fair algorithms to determine the competition results that might combine running, cycling and flying stages. Will humans compete against birds, or be combined in hybrid teams? Your technical platform should also include social media and live stream support to that can be scaled to global public participation such as refereeing, commentary, fan clubs or cash sponsorship.

Global Ground Truth

Client: Alec Christie (ICCCAD liaison) 

Local people in rural areas have valuable knowledge about climate change effects and adaptation, but it is not straightforward to integrate their observations into scientific evidence and policy frameworks. Your task is to create a trusted infrastructure that integrates reports from local people (including photos and voice recordings to be translated as necessary) into a secure evidence blockchain. The client tools must run on low-end android phones, and be accessible to people with low literacy. Processes for authentication and standardisation should respect traditional authority, cultures and rights of indigenous people, and expert advice will be available from the International Centre for Climate Change and Development (ICCCAD) in Dhaka.

Green Maps

Client: Pasquale Giovenale, IMC

In an effort to tackle climate change, we would like to reduce our carbon footprint. Therefore, we want to offer people insights into the carbon footprints of various transit options. Your task is to develop a system which finds routes using various transport types (cycling, car, train, plane) and computes the costs, travel time and carbon footprints of them. In the end, the user can then make an informed decision on what mode of transport to take.

Household Payment Pool

Client: Richard Watts, BigPay

Routine financial responsibilities can be more pleasant if a small amount of chance is involved, for example UK Premium Bonds, which distribute interest as monthly prizes with an average rate of return comparable to other government-backed investments. This project aims to create a personal finance game in which housemates (or larger groups) can exchange transactions. When you spend money, the transaction will be charged to some random member(s) of the group - possibly split. An important constraint is that you should never be charged more money than you would otherwise have spent in a given period (tuneable by either you or the group). Design suitable rules and build a UI and transaction recording system that allows people to play this game.

International Treasury Service

Client: Richard Watts, BigPay

Many companies have to work across currencies. BigPay operates in both Malaysia and Singapore, and has to use an external service provider to move money between the two. It would be more efficient to do this themselves, managing reserves of both currencies (MYR and SGD), setting an exchange rate, and executing and settling transfers taking at least a day to complete. Your job is to design a software system capable of managing and visualising these reserves. It should take a feed of the buy and sell rates available from partners at various distances in time (1 day, 7 day, 30 day) and export an API allowing users to give a target currency and amount and receive a price (and then confirm the transfer). As far as possible, the system should achieve optimal revenue. You will need to write a simulator for exchange rates and for customer behaviour.

Migration Simulation

Client: Steffen Oppel, RSPB 

The Yelkouan shearwater is a bird that migrates through the Bosphorus strait to reach the Mediterranean, but nobody knows how many there are. Your client has video of the migration as seen from different distances and angles. Unfortunately, computer vision algorithms can’t be trained without ground truth count of how many birds are there. You will create a CGI simulator of the flocking birds, including realistic atmosphere and viewing conditions, to generate simulated video with known ground truth bird numbers. The final stage is to train a machine learning system using your simulation to automatically retrieve the number of simulated birds, and then test it on a real migration video to yield the number of real birds.

Mobilising the University

Client: University Information Services

Many University systems used by Cambridge students can be extended using the new UIS API Gateway service, including location and access facilities that are currently delivered via the University Card, but which could in future be accessed via NFC authentication from the student’s phone. The goal of this project is to create a new student arrival experience, delivered via a phone app that integrates their official physical access to Cambridge with administrative onboarding and the online knowledge resources of the University.

Online Programming Game

Client: Lex van der Stoep, IMC

To teach beginning programmers problem solving and coding skills, we would like you to build an online programming game platform. The idea is for players to program bots which play a multi-player move-based game (e.g. Battleships, Snake, Pacman). The platform should expose a simple API through which the players can retrieve the current state of the game, and publish their next move. It should then display these moves into a visual representation of the game, allowing players to see how their bots are doing.

Personal Ambiguator

Client: Eleanor Drage, Centre for Gender Studies

Many areas of life such as housing, health, education and employment are subject to gender, racial and other biases that can be detected with machine learning classification systems. Your task is to identify potential bias in personal documents such as a CV, health record or university application, and then to use a generative neural network approach to tweak these documents (perhaps by adjusting text or selectively removing information) so that they cannot clearly be classified. The algorithm should be packaged in an interactive tool that allows people to write ambiguated personal documents, and also highlights and educates them about the sources of bias.

Reading the Leaves

Client: Julian Godding, Gardin

Plant leaves are anatomically complex but can be structured in ways to make a digital model of the plant. In this project your goal will be to use time-lapse photography to render a 3D model of plants growing in a vertical farm in real-time; mapping a grid-search of chlorophyll fluorescence measurements and displaying phenotype properties such as leaf area and weight, calibrated against ground truth measurements.

SMART Climate Goals

Client: Vedantha Kumar, Children's Investment Fund Foundation

The COP26 conference has highlighted the need for decisive action on climate change, both for governments and companies around the world. But which of these are just talking, and which are setting concrete goals? Goals can be evaluated using the SMART framework (https://bit.ly/3orDb9J - Specific, Measurable, Achievable, Relevant and Time-Bound). Create a system which can be used to analyse reports and statements, and identify climate change-related "goals" within them, and then evaluate these goals against the SMART criteria.

Smart Bins

Client: IMC

To help people become more sustainable, we would like to give them insights in their waste production. Your task is to develop a smart waste management system that keeps track of how you are disposing your waste. The idea being to make it more transparent so you can highlight areas to improve. The container would automatically weigh waste, and provide the user with insights through an app, so they see how their waste metrics compares to global and regional averages. A further feature could be to use image recognition to classify waste types, or to allow users to compete with friends on sustainability.

Social Media Wellbeing Filter

Client: Victoria Doerfer, Dovetailed

Social media is part of our everyday lives, but constant exposure can have negative effects - to the point where people remove themselves entirely from social networks. We want to make social media more positive for people, with a plugin that checks in with the user each day and can make suggestions or adjust feeds based on their mental state. How much negativity can they can deal with? Did they already have a bad day? Do they feel ready to see some bad news, but not too much? You might want to take on large platforms like Facebook or Twitter, or perhaps start with plugins or readers for a more specialist platform like Reddit.

Strawberry Fields

Client: Marc Jones, Antobot

Strawberry pickers currently spend a significant amount of time (approx. 20%) manually transporting trays to stations, generally at the end of the field. Autonomous logistics robots provide the opportunity to increase productivity by transporting the trays, thus reducing the demand on labour. Your challenge is to create a simulation of a typical harvesting scenario in ROS Gazebo and develop an optimised algorithm for efficient multiple robot path planning. Use this to recommend the minimum number of robots required (e.g. scheduling/logic) – too many robots will be complex and potentially too expensive, too few will create delays for the pickers and reduce efficiency. Based on the simulation results, you should identify required sensor and actuator technologies in order for the robot to optimally interact with human staff.

The Automatic Accountant

Client: Mark Parsons, Cambridge Enterprise

The Automatic Statistician is a system created by Zoubin Ghahramani and colleagues that looks for interesting patterns in data sets (by fitting Gaussian process models), and then automatically generates a research paper, including charts and natural language texts, describing the patterns that it finds. Financial accounting data is relatively constrained in its format, so it should be more straightforward to create an AI system that automatically generates professional-looking reports on a company’s performance, including capabilities for audit and compliance checking that work by recognising statistical anomalies. You will need to familiarise yourself with the research from www.automaticstatistician.com, and work with your client to design an original accounting application.

Urban Stories

Client: David Russell, The Fusion Works

The what3words service has become very successful, with its pitch that three random words can identify any location on the planet. If you’ve used it, you may have noticed that the random words are often strangely relevant to the Cambridge locations being specified - if not for one particular cell, then quite likely for the cell next to it. Your task is to use a combination of natural language processing and path optimisation methods to turn these into engaging stories, starting from the creative seed of places you have visited, or perhaps would like to visit.

Video Bones

Client: Phil Cambridge

The trombone is a beautiful instrument, at its best when viewed from multiple sides and heard from a distance, as in this video by Phil Cambridge: https://bit.ly/2YMweY6. Like Phil, we could all do with better tools for composing and publishing multitrack audio and video streams of the kind that became familiar during lockdown. Your task is to create a music-specific multitrack video editor with everything a brass player needs, perhaps incorporating solo and chorus, virtual camera pan and rotate of instruments and chamber groups, hip hop loops and video stutters, or other effects suited to your favourite music genres.

Wiki Editor-Editor

Client: George Gospodinov, Amazon

Wikipedia editors are among the few remaining public servants who we can rely on to distinguish between truth and lies. Many of the first generation of editors are still around, but the future of human knowledge might depend on newcomers acquiring better understanding of how to do this effectively. Your task is to analyse what makes a Wikipedia editor effective, using the WikiData corpus to identify which patterns of editing deal with controversies, result in lasting consensus, and so on. Using the results of this analysis, create an automated interactive training guide for new editors, giving them feedback on their own patterns of response, and helping them identify how they might deal with different cases.

Youth-led Future

Client: Oliver Williams, Curriculum for Life

Most young people (in wealthy countries) now have their education controlled via virtual learning environments. But preparation for the future will involve challenging the curriculum, not just accepting the status quo. Your task is to create an alternative mobile browser that school students can use to construct a customised view which collaboratively annotates, critiques, adjusts and (re)evaluates the front-end facilities of their school’s VLE. Just like the school strikes for climate, it shouldn’t be necessary to wait for permission, when creating youth-led alternatives to established skills and qualifications. This can’t be just another social media platform - Facebook is clearly the wrong model!