Research positions in machine learning, geophysical modelling, software engineering
Earth Rover Program (ERP) is a not-for-profit organization which started in November 2023, headquartered in the UK. Making novel use of cheap and dispatchable technologies, its purpose is to improve our knowledge of the soil, both enhancing scientific understanding and enabling farmers to raise fertility and crop production while reducing environmental impacts. This is a major initiative, supported by Bezos Earth Fund over the first 2.5 years, during which we will benchmark and image soils below-ground, improving our understanding of their properties on sites around the world. The initial research phase will enable us to refine the methodology before pursuing its further development and deployment. Our approach has the potential to help address one of humanity’s most urgent challenges: feeding the world within environmental limits.
ERP has been co-founded by soil scientist Simon Jeffery, Reader in Soil Ecology at Harper Adams University; George Monbiot, author of Regenesis and Honorary Fellow at Wolfson College, Oxford; seismologist Tarje Nissen-Meyer, Professor in Environmental Intelligence at the University of Exeter; and Katie Bradford, Operations Specialist working with climate organizations. The team seeks to expand the initiative with a range of further funders to support innovations in agriculture, food systems, soil science, and monitoring as a key solution to addressing the climate crisis.
Job description
To build the computational engine within ERP’s core science team, we are recruiting for three computational positions: 1) in machine learning to facilitate data fusion, augmentation, unsupervised classification of large, multimodal and complex datasets, time series analysis, (causal) inference, and 2) in geophysical modelling to assess soil properties for geophysical monitoring, including effective 2D and 3D forward modelling of seismic wave propagation at the scale of soils in (visco)elastic, acoustic, porous media, and applying inverse methods by using and further adapting state-of-the-art modelling tools from numerical methods and scientific machine learning. This modelling will be directly embedded in the workflows and platforms developed for observational data from geophysics and soil science, and intersect with the machine learning tasks from position 1). A third, initially part-time position can be filled in software engineering, to streamline our code workflows, incorporate the work done by the two above positions, as well as seismology and sensor engineering with onboard computing for scientific instruments, and to create a framework for efficient, decentralized databases.
The positions are offered with competitive salaries based on flexible work arrangements (regarding location, full-time/part-time) for highly motivated, independent, creative and engaged individuals with a strong cross-cutting collaborative work ethic and passion for the mission of ERP.
A prerequisite is evidence of independent research (PhD, publications) in respective fields, including but not limited to machine learning, data science (e.g. data ingestion/fusion, augmentation, inference, classification); seismology (e.g. computational modelling, numerical methods, seismic theory, inverse methods, machine learning methods (e.g. PINNs, FNOs) applied to partial differential equations, time series data). The software engineering position does not require a PhD, but extensive experience in coding with Python, C or similar languages, experiences with large databases, scientific datasets, and phone app development.
Place of work: ERP operates as a mostly remote team, with quarterly in-person meetups around the UK. The positions can be remote or (for positions 1) and 2) ) associated with the University of Exeter, in which case a further interview with Exeter will be undertaken. We are open to applications from candidates based anywhere, though flexibility may be required in core working hours to accommodate multiple time zones within the team. The roles include travel (in a most climate-friendly manner) to joint data analysis meetups in the UK and/or elsewhere around the world.
Start date/duration: We expect to onboard new roles between now and the end of 2024, in line with the development of the project. This phase of the funded project finishes in summer 2026, but we anticipate further expansion and longer-term funding.
Outcomes: The involvement in this research-intensive environment is expected to lead to high-profile publications, high visibility, global collaborations across disciplines, with equal footing in fundamental science and clear solution-oriented impact for public benefit.
Support: We seek independent researchers in the core team to collaborate closely, under the continuous guidance of science leads Prof. Nissen-Meyer (direct supervision for both research positions) and Dr Jeffery (joint supervision for software engineering position). Support funds exist for work-related travel and equipment. We strive to build a vigorous, diverse and inclusive environment with flat hierarchies and strong support on health and wellbeing matters. Remote work requires proactive communication skills, and the work environment will feel like somewhere between academic, non-profit and industrial research, with distinct timelines for producing outcomes.
Application process: We welcome applications from anywhere and anyone with relevant skills and backgrounds who feels they can positively contribute to and grow with the project.
Please submit your application until October 5, 2024 23:59 GMT via this form (https://docs.google.com/forms/d/e/1FAIpQLScMYLthRSurusd3Hot6B3xHGNx-VSit7FP5WvJLeNL50eVcJQ/viewform), which will require uploading the following documents:
• research statement including past experience and relevance for one of the posts (2 pages)
• CV, including email addresses for 3 referees familiar with your past research (2 pages)
• list of publications including open-source code, highlighting three most relevant projects (1 page)
• brief statement on logistics (1 page): where you would like to be based, when you could start, and whether full-time or part-time (this will not affect the judgement on your suitability for the technical aspects of the position, but a start date by December 2024 is desirable).
Text exceeding the above-stated lengths will not be considered.
Timeline: Upon receipt of applications, we will request letters of recommendation for shortlisted candidates, and perform virtual interviews by mid October. We anticipate extending offers of employment before late October for a start date as early as November 2024, and ideally no later than mid February 2025.
Questions? Can be submitted to the team using the form above, prior to job application.
FURTHER DETAILS ON EACH POST:
Machine learning: ERP’s central framework will be built upon machine learning, large and complex databases, data analysis and inference algorithms. This leads to a variety of data science tasks: Generating effective data ingestion from a diverse, heterogeneous and continuously increasing influx of large datasets, data fusion to allow for scientific inference across multimodal datasets, building modern, effective machine-learning algorithms for classification, Bayesian/causal inference, system discovery with complex ecosystems, uncertainty quantification, and effective interrogation of a vast database. This first post will devise the basic framework of the infrastructure in close cooperation with the geophysics and soil science teams, and will lead to a larger group focussed on building a virtual ERP environment.
Geophysical modelling: Complementing ongoing seismology efforts, this geophysical modelling post focusses on modelling seismic wavefields using numerical and machine-learning algorithms in 2D and 3D, as well as developing inversion and general inference frameworks. Input for the modelling will come from soil science, and close collaboration with sensor engineering on experimental design, and with machine learning and seismologists on incorporating modelling with observational data is expected. Much of the modelling will provide test beds for the computational infrastructure to be built, including inversion and testing multimodal datasets, while framing soils as complex ecosystems.
Software engineering: Consolidating ongoing efforts on internal code repositories, creating streamlined data acquisition workflows for hubs around the world, documented and reproducible dataset interrogation, incorporating codes from machine learning position, creating frameworks for large databases for complex, multimodal datasets, data security, basic knowledge of phone app development, collaborating with onboard computing/microengineering on scientific prototype sensors.
Location: , , GB
Offer Expires: 2025-03-25 17:35:43
Job Posting Language: en
LinkedIn: Apply Here