Principal Machine Learning Engineer

Accelerant

8d

About The Position

We're looking for someone to own how machine learning and AI run in production at Accelerant. You'll lead a small engineering function responsible for the platform our data scientists build on. That covers data and feature pipelines, training and inference services, deployment, monitoring, and the infrastructure behind our agentic AI work. You'll set the standards, coach the team, and be accountable for the whole thing staying up.

Much of the value in this role sits at the seams. Our machine learning systems are not an island. They need to exchange data and decisions with the wider Accelerant platform, with third-party providers, and with systems owned by other engineering teams. Designing those integrations, and building the working relationships with the people on the other side of them is closer to the centre of this job than any single piece of infrastructure.

We take the operational side seriously. We care about reproducibility, by which we mean knowing which data and which code produced any model currently making decisions. We care about training and serving computing features the same way, because the times they don't are the ones that hurt. We think about what we call the slow-label problem, where the ground truth on a claims or pricing model can arrive months or years after the prediction, and monitoring has to stay useful in the meantime. We have a bias toward dull, recoverable systems over clever ones that need someone awake to babysit them. If those are problems you've lived with rather than read about, we'd like to talk.

You'd be joining with some foundations already in place but without a decade of accumulated legacy to work around. There is meaningful scope to design the solution, and you'll be the person doing it.

Requirements

• Substantial experience running machine learning systems in production, including everything that happens after launch
• Strong engineering foundations. Python, infrastructure as code, containers and orchestration, and depth in at least one major cloud provider with sound instincts about cost and failure modes
• Data engineering capability, pipelines, orchestration, storage and access patterns, and enough SQL to hold your own in a warehouse
• A track record of integrating systems across organisational boundaries, including the part where you have to influence teams you don't manage
• Enough statistical literacy to have a real conversation with a data scientist about whether a model is working, and to stay skeptical when the dashboards say it is
• Experience leading or coaching engineers, plus judgement about which infrastructure will pay for itself and which is merely satisfying to build
• Willingness to work with LLMs and agentic AI as everyday tools, whatever your background is today
• The communication skills and credibility to be the person who says a system isn't ready

Nice To Haves

• Building infrastructure for LLM and agentic systems, including serving, orchestration, retrieval, caching, and keeping spend and latency under control at scale
• Regulated industries where model governance, explainability, and audit trails are requirements rather than aspirations
• Insurance or financial services, whether pricing, underwriting, claims, or portfolio management
• Having worked as a data scientist or predictive modeller at some point, or otherwise being fluent in how models are built and not only how they're shipped
• Internal platforms and tooling that other technical teams genuinely adopted, and a clear view of why they adopted them
• Real-time or streaming systems, feature stores, or high-throughput scoring

Responsibilities

• Owning the ML platform end to end, from data and feature pipelines through training infrastructure, model registry and lineage, inference services, and the deployment path between them
• Designing and building integrations with the wider Accelerant platform, third-party providers, and systems owned by other teams, working directly with those teams to get it right
• Making deployment routine rather than eventful. Versioning, staged rollout, rollback, and CI/CD for models and agents
• Building monitoring that separates data drift from pipeline breakage from genuine performance decay, and that stays informative when labels are delayed
• Standing up the infrastructure behind our agentic AI work, including orchestration, tool and API integration, retrieval and caching, and control of cost and latency
• Owning reliability, cost, and performance across ML workloads, from overnight batch scoring to low-latency services
• Building model governance and audit trails that satisfy regulators and internal risk committees without becoming a tax on design or delivery
• Leading and growing the function. Setting technical standards, coaching a small team, and partnering closely with the data scientists who depend on your work

Benefits

• Ownership of a function, the freedom to decide how it works, and a team of strong data scientists who need what you build
• Problems that span the full range, from overnight batch scoring to low-latency services to agentic systems, across more than 20 countries and 250 specialty products
• A collaborative group of people who enjoy solving difficult problems together

What This Job Offers

• Job Type: Full-time
• Career Level: Principal
• Education Level: No Education Listed