Analytic Learning Algorithm Research
We're building a system that represents domain knowledge as modular probabilistic models — making analysis rigorous and transparent. Users can connect these models flexibly into larger structures. The system enforces consistency across them, and propagates uncertainty through each step. Our first applications are in finance and scientific research, with use cases ranging from equity valuation and distress monitoring, to particle physics.
We are looking for full-time researchers to contribute to the development and analysis of our learning algorithms. You will work on interesting theoretical problems with immediate applicability to implementation of our system.
Our team works fully remotely, and mostly within the CET timezone.
Useful experience
Development of mathematical analysis methods, for example: optimal transport, information geometry, continuous optimization methods
Analysis of probabilistic graphical models, including factor graphs
Implementation of tractable density estimators (normalising flows, autoregressive density models, probabilistic circuits)
Translation between equational reasoning and code implementation
Mathematics, Computer Science, or Statistics advanced degree (with PhD or equivalent research experience)
Responsibilities
Develop numerical-analytical models of learning in our system
Connect our research to existing literature
Prove properties of algorithms and design experiments to validate results empirically
Leverage the expertise of other team members effectively
Write clean and well documented code
Help other team members to deliver on their goals
How we work
Hierarchical goals, not personal hierarchies: We organise around a transparent tree of goals and tasks. Every quarter we plan milestone goals, which branch down into smaller and smaller tasks. This tree is the foundation of how we organise, not a side tool.
Transparency: Everyone should have access to every opportunity in the team that they can realistically handle. All goals, tasks, and the reasoning behind them are visible to everyone.
Decisions become tasks: When something is discussed and decided, it gets captured as a task in the right place in the tree, so that nothing dissipates as hot air.
Written and asynchronous by default: We are fully remote and document our learnings in writing. Communication happens transparently in shared channels, not in private threads and one-on-ones.
Growing from leaf to tree: New joiners start from smaller leaves of the tree and work themselves up to ownership of larger branches as trust and understanding build. Teams form around topics and dissolve when the work is done; people move to where they are most useful.
On our website you can find more about our team and work culture, as well as example tasks that share some insight into the type of things team members are working on.
What we do: https://planting.space/
Ways of work: https://planting.space/org/
Team culture and example tasks: https://planting.space/joinus/
