ÂÜÀòÂÒÂ×

Who are we?

Oxa is enabling the transition to self-driving vehicles through an initial focus on the most commercially advanced sector; the autonomous shuttling of goods and people.We are home to some of the world’s leading experts on autonomous vehicles, creating solutions such as Oxa Driver, equipping vehicles with full self-driving functionality; Oxa MetaDriver, using Generative AI to accelerate and assure the safety of deployments; and Oxa Hub, a set of cloud-based offerings for autonomous fleet management. Our technology is being deployed across the UK and the U.S, and we’re partnering with a fast-growing ecosystem of operators, vehicle OEMs and equipment makers serving autonomous transportation globally as it advances.

Based in Oxford, and with offices in Canada and the U.S, Oxa was founded in 2014 and is  growing rapidly (350+ ‘Oxbots’ to date). Our purpose is to change the way the Earth moves, through an uncompromising focus on safety, efficiency and explainability of our AI approaches. The company has attracted $225 million from leading investors so far, with $140 million raised in the last Series C funding round in January 2023.

Your Team

The Data Capture team aims to collect and curate the datasets which contain the variation and volume needed to robustly train and certify Oxa’s autonomous vehicle technology.  We aim to help define the problem of autonomous driving in data to enable easier adaptation of Oxa’s technology to new areas. 

Data Capture sits within the MetaDriver product group which is a suite of tools that combines generative AI, digital twins and simulation to accelerate machine learning and testing of self-driving technology before and during real-world use. Data as with all AI and machine learning forms a crucial foundation for this as well as for all of Oxa’s technology.

Modern AI/ML relies on data. To solve a problem and more importantly prove that you have solved it you must define that problem in data! In order to achieve Oxa's aim of Universal Autonomy, we must collect quality, varied data in volume. This is key not only to train the best autonomous vehicle systems but enables robust certification of their safety. 

Your Role

Our Data Scientists role is to optimise the deployment of a suite of data collection vehicles, developing novel ways to collect the most varied datasets, such as using automated route planning. You will be at the forefront of automating data curation pipelines using the latest machine learning techniques and optimising the use of manual labelling. Continually researching state-of-the-art data representations with self-supervised learning to explore and understand large datasets at scale and building certifiable dataset benchmarks which robustly prove different components of the autonomous technology stack.

Your Responsibilities

• Vehicle Data Collection Improvements: Investigate ways of improving data capture within the vehicle. e.g. collecting voice recordings from the driver, collecting additional metadata and onboard processing of data to provide automated instructions to the driver.
• Vehicle Logistics: Ensure that there is a clear plan for where and when the vehicles should drive and  collect data. Organisation of drivers, GCA copying etc. Ensure adherence to regulations etc.
• Automated Data Labelling: Building automated data labelling pipelines to provide richest possible data labels which are ready for training.
• Data Representation Learning: Researching the best representations of data to enable search, cleaning benchmark creation and data gap analysis.
• Data Cleaning: Automatically clean data of any unwanted anomalies, understanding the cause of these anomalies and correcting the data collection process if needed.
• Data Search: Enable searching of data to allow teams to easily find the data they need to solve a task. Hard example mining.
• Data Gap Analysis: Automate the discovery of features and events that we have limited or no examples of. Develop methods for discovering the most efficient options for collecting new data to fill the identified gaps.
• Data Benchmark Creation: Create robust benchmarks that test the capability of various algorithms. Different benchmarks will be needed to test different techniques and scenarios.
• Offline Assurance and Benchmark Evaluation Tools: Develop software that enables standardisation of how we report performance of techniques on the various benchmarks.
• Contributing to high-quality, written, technical reports and visual media which represent progress and accomplishments in the company data agenda