This job might no longer be available.
7 months ago
Improbable Defence, building on the backbone of SpatialOS, has combined world class scientific modelling, market leading AI, mission specific user interfaces and a uniquely flexible and secure deployment model to create a powerful simulation platform tailored to the needs of the war fighter.
Our mission? To enable the most realistic and effective military simulations ever experienced, making defence users more effective on operations and decreasing the cost of military preparedness.
is to help us achieve step changes in capability for simulations of unprecedented realism and complexity, for the purpose of ‘in-silico’ experimentation with the effects of aggressive and defensive action on socio-technical systems: the intertwined systems that make up the environment and society. Via R&D in machine learning, statistical inference and computer science, we produce novel methodology and tooling for modellers of complex systems.
We are a team of experienced researchers and software engineers interested in removing computational and information bottlenecks in the design and validation of models of complex systems, typically aimed at counterfactual analysis and simulation.
We aim to ensure that models of complex systems are sound, statistically calibrated against real-world evidence, and built in a modular fashion to ensure that more complex models can be synthesized by combining models of simpler components. We are also interested in policy optimisation and reinforcement learning strategies for multiple autonomous agents operating within this synthetic environment. This puts our work at a crossroads between agent-based modelling, multi-agent systems and complex systems.
We collaborate with some of the leading academic research groups in the UK. We take inspiration from the success of general-purpose tuning algorithms and resulting frameworks for deep learning (e.g., Tensorflow) and hierarchical Bayesian models (e.g., Stan or Pyro), and aspire to build analogous tools for modelling complex agent systems. Our prior work includes an in-house probabilistic programming language, Keanu.
Areas for Impact
- Become rapidly well-versed in the state-of-the-art in agent-based models and multi-agent systems as captured in the literature and open-source projects.
- To pick up statistical, machine learning or mathematical modelling techniques as required independently from the literature, and be able to implement them in a scientific programming language (R, Python, Julia), or, in the case of programmers with experience in lower-level languages, using mathematical/scientific libraries.
- To pursue independent research in a prescribed strategic direction. This assumes an ability to translate ambitious but often vague requirements into concrete goals, and work under tight time frames without compromising on scientific integrity.
- To communicate both advances and failures of research initiatives to other researchers as well as non-technical audiences, with confidence backed by data.
- To transfer scientific know-how so as to empower more junior researchers in the team, as well as scientific modellers and engineers in the broader division.
- To deeply understand and take ownership of the vision for next-generation software for planning and training for defence and security - to save lives and public funds. This entails parsing sometimes arcane material on wargaming, military simulation and other niche areas, so as to detect room for improvement.
- To experimentally validate novel statistical procedures in the absence of theory.
- To work in teams comprising individuals with varied technical backgrounds.
- To care about maintaining and improving the status of our workplace as an inclusive, diverse working environment.
We'd like to hear from you if you identify with the following:
- PhD in statistics, mathematics or a related area. Background in computer science with training in stats and/or machine learning are also welcome.
- Expertise in scientific programming in a higher-level language for rapid prototyping. For example, statistical and scientific computing in Python, R, or Julia. Adept scientific programmers in other languages are still encouraged to apply.
- Experience independently implementing statistical algorithms from scratch (i.e., from a mathematical description).
- Evidence of ability to pursue independent research (e.g., first-author papers or postdoc experience).
- Demonstrable interest in gaming, simulation, defence, probabilistic modelling; or all of the above.
While we think the above experience could be important, we’re keen to hear from people that believe they have valuable experience to bring to the role. If you identify with the team and mission, but not all of our requirements, then please still apply!!
The best ideas are often the least expected and require new ways of thinking; that’s why our teams at Improbable are made up of an incredible range of talented people. Improbable is proud to be an equal opportunity employer. We do not discriminate based on race, ethnicity, colour, ancestry, national origin, religion, sex, sexual orientation, gender identity, age, disability, veteran status, genetic information, marital status or any other legally protected status.