Projects
Join a project to create novel research or commercial products.
We run challenging projects throughout the year that are proposed by experts in their fields (Researchers/PhDs/Entrepreneurs). If you are interesting in joining a project team, please apply using the link below.
Apply for a project
If you want to propose your own project idea. Please fill in the project proposal form. We will help you build a team and provide the resources to get it started.
Propose a project
If you have any administrative questions for the projects (e.g. joining a running project if you are SUPER keen) contact us on facebook or email, rather than contacting supervisors.
- Research
- Commercial
- JumpStart
- Consulting
New Projects [accepting applications]
-
RapiDx: (summer internship) PCR assay development - experiment and develop new rapid diagnostic assays for a wide range of diseases. Learn to operate plasmonic PCR machines developed by the team at RapidX Bio and engage with experimental planning, data collection and analysis and most importantly enjoy the process of inventing life-saving tools that are 20x better than what’s out there today.
RapiDx is building the future of healthcare, with an optical plasmonic PCR (polymerase chain reaction) device that can help detect microbial diseases in minutes.
-
Biomr: leveraging ML algorithms to discover and develop better biopolymers, particularly polysaccharides, as alternatives to restricted or banned ingredients in consumer products.
Biomr is pioneering ML-aided sustainable materials discovery, addressing the challenges faced by consumer goods companies in the wake of increasing ingredient bans, particularly in the EU. Biomr utilizes classic machine learning (ML) with state-of-the-art artificial intelligence language models to speed up the identification and development of biobased material alternatives with high performance.
Active Projects [closed applications]
-
Playfair (Hedge Fund API): construct Financial Signal dataframe API for quant-funds using proprietary social media signal data derived by our Playfair AI Models.
At Playfair, we use ML system to classify the opinions of millions of traders on social media. In doing so, we find the world’s best as we track their performance against the market. This data is used to construct the world’s first social finance ranking system.
-
Playfair (Algorithmic portfolio): create trading algorithms for a $300k USD portfolio in Scotiabank’s Capital Market’s group using buy/sell signals from the best traders on social media.
At Playfair, we use an ML system to analyze the opinions of millions of traders on social media. In doing so, we find the world’s best as we track their performance against the market. This data is used to construct the world’s first social finance ranking system we sell to large hedge funds via API.
-
Caspa AI (AI option): developing AI methods to quickly, easily, and accurately generate custom product images and marketing content for ecommerce sellers.
Caspa AI is a generative AI startup developing software that lets ecommerce businesses create realistic and customized product images.
-
Caspa AI (marketing option): developing effective marketing strategies, everything from social media to SEO. This is a perfect for anyone who wants to get real-world experience in marketing, growth and business.
Caspa AI is a generative AI startup developing software that lets ecommerce businesses create realistic and customized product images.
-
Biomr: (option 1) materials-centric simulation and modeling to develop revolutionary plastic-free coatings; (option 2) creating advanced sorting algorithms to facilitate textile recycling and circular fashion models.
At Biomr, we firmly believe that we are on the brink of a materials science revolution, aided by computation and the growing demand for natural materials driven by regulations. We are a small but ambitious team dedicated to pushing the boundaries of materials science.
-
ThinkLabsAI: building a website and GPT-powered conversational chatbots to provide affordable, personalized career guidance and academic support.
ThinkLabsAI is a Generative AI startup. We seek to build a platform that provides personalized, unbiased, affordable, 24x7, instant, yet human-like support for students at all career stages in higher education.
-
ReplyGen (B2B SaaS Generative AI): building a website and generative AI models which will allow customers to make custom marketing campaigns.
ReplyGen is a Sales and marketing + Generative AI startup. What we are trying to solve is to use generative images and video AI to create highly customisable marketing solutions easy to build without much technical expertise on the company's end. We also want to make it cheaper and less time intensive to build new marketing and sales campaigns.
-
Playfair Technologies Inc. : Build an ML model to pick up market sentiment based on technical analysis images posted by the world's best traders.
At Playfair, we see value in finding the value of consolidating a diverse range of valuable insights from a variety of data sources, in particular, technical analysis (TA) reports in financial markets to best predict accurately the direction of markets. We hoped to build a Computer Vision model that can organise such data provided by the world's best traders to make good trading strategies and hence determining the change in market sentiments.
-
NLP Interpretability: How does GPT-2 understand language?
In this project, what we wish to study in depth is to understand how Language Machine Learning models are able to construct meaning from arbitrary language characters and decipher complicated language tasks. In particular, what we would like to achieve is to build a framework in a way that such language ML models can create new formatting, documentation, text and much more versatile uses in the world of Natural Language processing
-
Ethēcos: Development of an adaptive platform from a range of manual input and biomarker health parameter data.
At the heart of health-tech, nutrition, and education, Ethēcos believes in the immense potential in extracting insights from biological marker information for consumers to make informed decisions to staying healthy or simply to be the best version of themselves in a simple and comprehensive way in the context of education or self-awareness. In this respect, our ambition is to build a mobile app product that collects such data of our bio-marker parameters and evaluate what this means to our health, progression in our education, etc, thereby giving consumers vital information to make small changes in their lifestyle.
-
Kanjo Health Project: Developing cognitive, linguistic and wellbeing biomarkers and biometrics from a range of device interaction and sensor data
Almost every parent wishes they understood their children better. This manifests in a broad range of questions and concerns, such as: when is the best time for my child to study, are they getting enough sleep, why are they currently anxious, are their language skills developing well, am I doing an okay job as a parent? We aim to aid parents with all these questions; helping to build a generation of more emotionally resilient, engaged and happier children - and their parents
-
Hackbridge Device Innovation Addenbrokes Medical Engineering Project
For this project, we need to design a patentable one-way locking (unable to unlock once applied) lid container mechanism. This is important in the fields of medical science to ensure the quality of the samples collected is not compromised. In addition, the containers have to be sturdy enough to prevent common damages or shock applied to the casing, sample, etc. Design features are outlined in the document
-
NLP based Project by Opply in faciliating Effective Supply Chains
Implementing an algorithm to establish defined groups of suppliers relevant, collecting a wide scope of data to ensure enough information is collected based on supply chains as well as bridging the gaps between product quality and quality of data on product supplies.
-
Segmentation Network to Locate and Predict Damage to Bikes
For teams managing shared bikes and e-scooters, they currently have no way to see or predict what types of issues will happen. This leads to endless support tasks and frustration for riders. We aim to solve this by building machine learning models to input an image and output a prediction of the type of damage and location of damage on the vehicle.
-
ML-Enhanced Computer Vision Change Detector System for Satellite Images Analysis
For this project, we are building upon the work we have carried out in the past semester of the Hack bridge project, generate detailed performance metrics of the change detector and building height detector, and test it in a wide range of use cases (Oil spills, vegetation encroachment, building damage etc). We’ll also want to have a standardized documentation in the form of Confluence wikis for all the processing blocks developed and if time permits, to file a patent
-
Embedded lending for marketplaces and vertical SAAS platforms
Implement a credit decisioning platform to underwrite loans to the businesses that are selling on an online platform, using their platform transaction data.
-
Perform unsupervised clustering of documents to detect similarities.
Explore datasets of pdf documents (images) using unsupervised learning techniques like clustering and anomaly detection methods.
-
“Computer! Sort Out my PDF”: Build an ML model that identifies the structure of a pdf containing multiple documents
Extend our ML service which aims to classify and paginate documents in a PDF using computer vision and natural language processing techniques.
-
[Flagship Project - Convertible Internship Available] Automatic Tile Grouting Robot
Fabrica.AI builds automatic tile grouting robots and eventually automatic tiling robots. Our robot is meant to automate the whole tile-grouting process. The robot will help boost productivity and help our clients return their investments in a year or two when purchasing our robots.
-
Dashboard for a Solar-Electric Vehicle
The project is a user interface for an electric vehicle that displays vehicle information such as speed, battery charge, etc.
-
[Flagship Project - Convertible Internship Available] Developing a calendar-based nudge tool helps senior leaders make small but impactful changes in behaviour and communication at work.
Use Machine Learning to create a tool that reads data from a calendar and create unique nudges. The nudges should be relevant to their next meeting; fitting within the natural flow of work. A nudge follows a learning curriculum which uses predictive modelling to calibrate present, priority and missing skills.
-
Develop a demand forecasting time series models and MIP optimisers.
PreWarp is focused on helping fashion brands earn more and waste less. We are building demand forecasting and revenue-maximising models using data science and mathematical optimisation algorithms.
-
Dissecting deep learning for systematic generalisation
We carry out an empirical study on the effective computational power of deep learning methods (transformers in particular) for systematic generalisation.
-
Head Circumference Measuring App
The aim of this study is to devise and validate a mobile app utilising modern smartphone camera technology to allow parents and caregivers to reliably measure infant HC. An accurate head circumference (HC) measurement is essential when clinically evaluating a neonate or infant. HC outside the normal range may indicate a disorder of brain development, hydrocephalus, or an intracranial mass lesion. A project hosted in collaboration with Addenbrooke's.
-
[Flagship Project - Convertible Internship Available] Developing a calendar-based nudge tool helps senior leaders make small but impactful changes in behaviour and communication at work.
Use Machine Learning to create a tool that reads data from a calendar and create unique nudges. The nudges should be relevant to their next meeting; fitting within the natural flow of work. A nudge follows a learning curriculum which uses predictive modelling to calibrate present, priority and missing skills.
-
ML-Enhanced Computer Vision Change Detector System for Satellite Images Analysis
To fight climate change by inspect the world’s ocean to detect oil spills and land for damage estimation following natural disasters.
-
Developing an integrated, personalized medicine platform using digital twins extrapolated from real-patient data
By extracting each patient’s unique biological signature, we aim to redefine the dynamic between patient, drug and treatment. Analogous to online ads targeted to your online identity, Aiina aims to accelerate therapeutic and clinical solutions to your unique biological signature.
-
Automatic Rowing Boat Steering System
[missing app developer] Create a low cost auto pilot aiming system for small boats and rowing boats using the computing power on a mobile phone.
-
Open Source DNA Imaging System with Raspberry Pi - In Collaboration With
Design and build a DNA imaging system using a Raspberry Pi Camera for use in resource-constrained labs like Biomakespace and our partner labs in Africa.
-
Deep learning for data validation in industrial formulations and processes
Deep neural networks are implemented to automatically validate and extract inter-property correlations from complex and expensive experimental data. This data is typically recorded at source in a non-systematic way that is prone to error and leads to delays in R&D pipelines. Our deep learning tool, Alchemite, can uniquely interpret numerical data in a way that is consistent across all domains.
-
Applying Explainable AI to Deep Learning across Industries
Untangle AI has a experimental SDK/toolkit for explaining deep learning models. This project involves testing and improving the toolkit applied to problems such as Breast Cancer and Object Detection.
-
Create Remixes of Musical Pieces Using Deep Learning
Analyze audio and extract central features using deep learning applied audio and forge remixes by mixing features in complex ways while keeping the identity of the music.
-
App-Building for those Affected by Allergy-Like MAST - Collaboration with ITIC
Create a platform for patients suffering from MAST and build a profile on what is causing their allergy-like symptoms.
-
Accelerating Homomorphically-Encrypted Neural Networks on an FPGA
This project seeks to implement an accelerator for homomorphically-encrypted neural networks on an FPGA. We will be competing in the Xilinx Open Hardware 2020 competition, in the Accelerator category.
-
Building an impact metric through quantitative analysis
[Consulting] InsideUni is a charity that aims to make Oxbridge accessible to everyone. The charity is currently in need of sophisticated analysis of vast datasets to understand the impact of its website and other channels on various stakeholders.
-
Customer and Volunteer Profiling through analytics
[Consulting] This is a LSC Hackbridge project with SolidariTee that aims to develop a comprehensive and holistic profile of their customers and volunteers to better inform any strategic decisions such as their product lines, community engagement models and marketing campaigns.
-
Open Source Protein Electrophoresis - In Collaboration With
Design and build an open source, safer, affordable protein gel electrophoresis system for use in resource-constrained labs like Biomakespace and our partner labs in Africa.
-
Open Source UV Spectrophotometer - In Collaboration With
Design and build an open source UV spectrophotometer for quantifying DNA and proteins for use in resource-constrained labs like Biomakespace and our partner labs in Africa.
-
Reinforcement learning with a robotic arm- In Collaboration With
Detailed task to be specified in the first meeting, options are robot playing a game like Jenga or Mikado or learning to grasp objects with a passive anthromorphic hand. Details of one option in proposal.
-
Generative models for 2D and 3D cancer imaging
Developing 2D or 3D GANs to generate medical scans of cancer as a means of data augmentation for cancer segmentation models.
-
Sensor Fusion for Automated Driving
Reduce computational requirement for multi-level sensor fusion - Implementing full-stack algorithms to allow autonomous navigation of vehicles in virtual hardware environment
-
Dynamic Vehicle Weight Sensors
The aim of the project is to develop a system to accurately measure the weight of each axle of a lorry as it drives past at highway speed.
Past Projects
-
Application to Resolve the Knowing Doing Gap
We’re testing a product which helps founders, execs and managers of fast paced companies resolve the knowing-doing gap. The Knowing-Doing Gap is when we are busy and stressed, we forget to apply the behaviours, we know will optimise team performance, and default to old ways.
-
BitFear: fear indeces for cryptocurrencies
Develop a volatility index for cryptocurrencies
-
Virtual Metrology on Plasma Etching
Build a machine learning algorithm for predicting the result of a quality test after a plasma etching process taking as input machine variables (temperature, pressure,...) recorded during the process.
-
Matching and Optimisation Algorithm on Pooled Data
Designing and implementing a data-matching and optimisation algorithm, on encrypted financial and personal data. The project also includes transforming the input data (and as a bonus, integrating your work into our existing user flow).
-
NLP for News Topic Classification
Natural Language Processing classification algorithm for multi-media news content.
-
HelioRec
An offshore floating solar power plant manufactured from recycled plastic: sustainable, scalable, rapid-to-deploy, reliable and cost-effective source of power generation..
-
Designing an Automated Marine Steering System
Designing a low cost auto pilot aiming system for small boats and rowing boats using the computing power on a mobile phone. Proposed by the Founder of Hawk-eye - Paul Hawkins
-
NLP for News Clustering
Natural Language Processing clustering algorithm for multi-media news content.
-
Combating Food Waste With Computer Vision
Developing a computer vision algorithm to recognise food ingredients, used for recommending recipes for leftover food.
-
Image Classification Using a Sequential Attention Mechanism
Developing a new image classification method, in which a neural network looks at an image multiple times before making a decision
-
Investigating The Gravitational Collapse of Planetary Atmospheres
Study the collapse of planetary atmospheres via hydrodynamic simulations.
-
Video Generation via Generative Adversarial Networks
Producing realistic video sequences using GANs
-
Hypervault - Blockchain-based Secure Resource Access System
Designing a secure resource access system that makes it easy to safely share confidential information
-
6DOF Object Pose Detection from Monocular Images
Re-implementation and possibly extension of a neural network architecture for 6DOF object detection from monocular images.
-
Beta Go - A Reimplementation of AlphaGo Zero
Reimplementing a scaled down version of Deepmind’s AlphaGo Zero, playing on a 5x5 board
-
Mandarin Tone Trainer
Using machine learning to develop a system that can recognise and correct Mandarin tones, for people learning the language