The role
We are looking for a machine learning engineer to keep that pricing engine running reliably in
production. Your focus is the machinery around the models: the pipelines that train and run them, the
infrastructure they run on, and the monitoring that tells us when a model or a daily pricing run has
gone wrong.
This is a hands-on contributor role in a small, distributed team. You will pick up well-defined pieces of
work and see them through, with growing ownership of the operational side of the engine as you learn
the domain. Deep modelling expertise is not expected on day one, though you will be working
alongside the models every day.
Monitoring and observability is an area we are actively investing in, and it is where this role will have
the most immediate impact.
Tech stack
Area Technologies
Backend Python, FastAPI, Node.js, GraphQL, REST APIs · Rust an advantage
Cloud & DevOps Docker, AWS (EC2, S3, Lambda and similar), Git, CI/CD, infrastructure as

code (Terraform / OpenTofu)

Testing pytest, Playwright and similar automated testing frameworks
Data Columnar dataframes and Parquet, PostgreSQL, SQL
Orchestration Scheduled and event-driven pipelines for model training and batch inference
Model operations Model registry and versioned artefacts, reproducible training runs,

automated validation before release

Monitoring Pipeline and job observability, model performance and data drift tracking,

product analytics

AI tooling Claude Code, Cursor and similar AI-assisted development tools

Area Technologies
Analysis Notebooks and Python visualisation libraries
What you will be doing
Building and maintaining the containerised pipelines that train models and produce daily price
recommendations, along with the AWS scheduling and queueing that drives them.
Watching daily pricing runs across every customer: catching failed or degraded runs, working
through dead-letter queues, and getting a run back on track before it affects a customer's prices.
Building out model performance monitoring — tracking prediction accuracy over time, input data
drift, and how recommendations compare to what was actually sold.
Extending the automated checks that stop a bad training run reaching production.
Maintaining the model registry: versioned artefacts, and the assignments that decide which model
version serves which customer.
Owning CI/CD for the ML repository — build, test, package and release across a monorepo of
independently versioned packages.
Writing and maintaining the Terraform / OpenTofu that defines the ML infrastructure.
Supporting the team's modelling work: preparing data, reproducing results, and productionising
experiments once they graduate.
Improving the runtime and cost of training and inference jobs.
Essential experience
Commercial experience as a machine learning engineer, MLOps engineer, data engineer, or
backend engineer working closely with ML systems in production.
Strong Python · modern, type-annotated code, properly packaged and tested.
Experience running things in production on AWS: containers, scheduled or event-driven jobs,
queues, and object storage.
Infrastructure as code (Terraform, OpenTofu or CDK) and CI/CD pipelines.
Practical monitoring and observability experience — logs, metrics, alerting, and the judgement to
know what is worth alerting on.
Solid SQL and relational data skills.
Automated testing as a normal part of delivery, not an afterthought.
Working fluently with AI coding tools such as Claude Code or Cursor as part of your day-to-day
delivery.
Clear written communication, and comfort working in a distributed team where technical decisions
are documented and debated in writing.
Comfortable working to a direction set by someone else, and confident asking for it when it is not
clear.

Desirable
Model performance monitoring and drift detection in production.
Experience with an ML platform or model registry (MLflow, SageMaker, Metaflow, Kubeflow) —
ours is in-house, so the concepts transfer more than the tool.
Familiarity with gradient boosting and tabular machine learning; any exposure to causal inference
or sequential decision making is a bonus.
Product analytics tooling (PostHog, Amplitude, Mixpanel or similar), and using product usage data
alongside model metrics to understand how recommendations are actually being used.
Comfort reasoning about memory and throughput on large datasets.
Multi-tenant SaaS architecture, including tenant isolation and per-customer model versioning.
Dynamic pricing, revenue management, yield management or e-commerce pricing experience.
Experience in the finance or fintech sector, particularly where automated decisions carry
commercial consequences and need to be auditable.
Rust, which we are progressively adopting for backend services.