Mid-Senior Machine Learning Engineer

We’re looking for a mid-to-senior level Machine Learning Engineer to build, deploy, and scale the models and systems that power Up. This includes creating intelligent features in the Up app, as well as developing systems to automate internal processes, increase the efficiency of customer support, and forecast business impact. You’ll be joining a small team of Data and ML Engineers, tasked not just with using our data platform, but expanding its capabilities to encompass ML training, inferencing, and model deployment. Our focus is on using our rich data to build intelligent systems that will improve the financial lives of our customers and enhance how we serve them.

This is a rare opportunity to join our small but growing Data & ML team and have an outsized impact on the future of Up, Australia’s highest-rated banking app. Backed by Bendigo Bank, we operate with the speed of a startup and the scale of an established institution.

The role’s impact goes far beyond training models in a notebook. You’ll own and automate the end-to-end lifecycle of ML models, taking them from initial concept through to production systems that serve millions of customers:

• You'll play a key role in moving our ML capabilities forward, helping us transition from infrequent, manual training to fully automated, continuously monitored systems.
• You’ll be pivotal in the development of our MLOps practices and tooling, ensuring our models are robust, reliable, and easy to maintain.
• You will be at the forefront of our natural language processing initiatives, deriving value from unstructured text data. This includes building and maintaining an embeddings store, training classification models, and fine-tuning transformers.
• You’ll help scale our ML-powered features to cater for the ever-growing customer base of Up. Features such as intelligent customer chat routing and fraud detections.
• You will play a vital role in the entire lifecycle of our ML projects. You can expect to take on significant responsibility in designing, building, and delivering solutions to some of our most interesting problems.
• When we have a gnarly problem like how to serve a model in real-time with low latency, or how to detect and mitigate model drift, we’ll look to you to help design and implement a robust solution.

Experience in some or all of the tools that we use will be beneficial, but we can give you the space to learn anything you don’t know already:

• Google Cloud is our cloud platform of choice. Our data lives in BigQuery and Postgres, and our applications run on Kubernetes. All of that is managed with Terraform.
• Dagster orchestrates our data ELT, and we are expanding its use to manage the full MLOps lifecycle for our models — from training, evaluation, and experimentation, all the way to deployment.
• Most of our code is written in Python and SQL. Most of our custom models are written with PyTorch, though we use off-the-shelf models in Google Cloud to fill in the gaps when a custom build isn't needed.
• Regardless of language or framework, we’re looking for someone with strong software engineering fundamentals who can apply them to building robust and maintainable ML systems.

• Experience managing the full lifecycle of machine learning models: from data acquisition and feature engineering to training, validation, deployment, and monitoring in a production environment.
• An opinion on what makes a good ML system, particularly when it comes to reproducibility, testing, and maintainability. Good communication skills will be key.
• Experience in building and operating the infrastructure that supports ML systems. You're comfortable thinking about not just the model, but the entire system around it, from CI/CD pipelines to serving infrastructure. Experience with NLP, including embeddings and tuning transformers, is a big plus.
• An intrinsic motivation to collaborate and share knowledge with the engineers around you, helping us all raise the bar for how we build intelligent systems.
• An understanding of how ML can be used beyond just prediction; to automate processes across the organisation, to develop new personalised features for Up’s customers, and to glean insight into the future.

#LI-Hybrid