Job Specifications
About Moneybox
At Moneybox, our mission is to give everyone the means to get more out of life. We're guided by our belief that wealth isn't about the money, it's about the means to more - more freedom, opportunities, possibilities, and peace of mind. Moneybox is an award-winning wealth management platform, helping over one and a half million people build wealth throughout their lives, whether they’re saving and investing, buying their first home, or planning for retirement.
Job Brief
We are building Aurora, an AI system designed to guide customers toward better financial outcomes. The core technical challenge is hard: given a customer with incomplete, uncertain information about their own financial situation and goals, how do you reliably converge on the right guidance - at scale, in a regulated environment, with decisions that must be auditable and traceable?
This breaks into several non-trivial subproblems. How do you efficiently resolve uncertainty about customer state through active information gathering, asking the right question at the right moment rather than exhausting the user? How do you translate natural language policy and regulatory constraints into formal optimisation logic that is both correct and inspectable?
How do you orchestrate learned and symbolic components such that the overall system behaves reliably, degrades gracefully, and can be reasoned about by humans? How do you do all of that without paying the engineering overhead on the expert parts of the system?
We have working hypotheses and committed architectural directions on all of these. We also change our minds quickly when presented with strong arguments or new evidence. If you think we’re wrong about something, we want to know.
We host most of our models internally. We develop using Databricks@Azure, and we deploy through Databricks, or directly on Azure Kubernetes Service (AKS).
This is the foremost research position in the ML team. You will report directly to the Director of AI and Decision Intelligence and work alongside a principal data scientist, senior ML engineer, senior data scientist, and two ML engineers.
What You’ll Do
You will work with other ML researchers, data scientists and ML engineers to:
Propose and prototype architectures that address our core problem set, with clear-eyed assessments of engineering complexity and scalability tradeoffs
Code proofs of concept to validate hypotheses and test the limits of theoretical approaches, grounded in empirical reality rather than theory alone
Serve as the team’s research lead - setting the intellectual agenda, running literature reviews, and keeping the team calibrated to what is actually state of the art versus what is hype
Challenge existing architectural decisions, including ones we are currently confident about
Input information to the Director of AI and Decision Intelligence and the wider AI team for decisions on objective functions, data strategy, and content strategy to ensure long-term coherence between research direction and overall system goals
Collaborate with academic partners where relevant and possible, with scope to contribute to publishable work emerging from applied research
Who You Are
You think in systems - you can hold the interaction between components in your head, reason about failure modes, and identify where theoretical elegance will break against production constraints
You have an optimisation mindset, not just in the technical sense but in how you approach problems generally - you look for the lever, not the brute force solution
You are comfortable with high autonomy and wide scope in a fast-paced environment
You deliver great results while managing your own pace sustainably
You read the literature seriously, have opinions about it, are willing to defend them, and willing to update them
You are not scared of ambiguity and thrive more when the problem is harder than when the solution space is clear
Experience And Skills – Essential
3+ years in an ML research or engineering role with meaningful exposure to text generation, agentic systems, or symbolic reasoning (one of is fine) - or equivalent academic experience with real applied components
Demonstrated ability to prototype rapidly and evaluate results honestly
Knowledge of applied machine learning, model tuning and model evaluation
Knowledge of the latest approaches in generative AI, including SoTA models
Familiarity with open-source LLM ecosystems and cloud infrastructure sufficient to bootstrap independently
Experience and skills – not essential for the role, but will be counted as a plus
Experience combining deep learning with formal or symbolic systems - this is the closest to what we are building and the strongest signal we can receive from a candidate
Familiarity with probabilistic graphical models or decision-theoretic frameworks
Experience in regulated or high-stakes deployment environments
Working in a scaled B2C environment
Experience deploying using any of: Databricks,
About the Company
At Moneybox, we help you turn your money into something greater.
Millions of us want to achieve more with our money. But whether we’re looking to save for a rainy day, grow our money, buy a home, or even build a retirement fund, we leave it at the bottom of our to-do lists because we're not sure how to get started.
This isn’t surprising. We aren’t taught about financial planning at school, the wealth industry was built to serve a minority, and it often feels like banks don’t care about helping us achieve outcomes. To top i...
Know more