IoT Data & AI Engineer

As Senior IoT Data & AI Engineer, you will convert proven analysis into running systems. Working alongside our IoT data analyst, who holds the domain knowledge and customer relationships, you will build the data platform, the predictive models and the automated reporting that allow insight to be produced repeatedly rather than by hand. The first priority is early failure detection. We have committed to delivering it for a major customer, and getting it into service - then proving how accurate it is - is your first quarter.

This is a hands-on role in a deliberately small team. You will work directly with the CTO and with our IoT data analyst, who is your closest working partner: they define what good looks like and validate that what you build is right; you make it repeatable. You will also collaborate with the wider Airedale group data team on shared standards as that function develops.

You will not be required to run the monthly customer reporting meetings - the analyst leads those - but you should expect to join customers periodically and for technical discussions. Engineers who never meet the customer build the wrong thing.

• Take our anomaly detection work from proof of concept into production across the estate, moving from retrospective analysis to live early warning.
• Build the validation loop that joins every alert to the engineer's finding, so detection accuracy is measured rather than asserted.
• Develop the failure-signature library from telemetry crossed with twenty years of service history - the asset that makes our predictions defensible and that no competitor can replicate.
• Tune alerting continuously so it stays actionable and does not become noise.
• Own the flow from Things Board, currently on managed hosting, into Microsoft Fabric - streaming ingest, full-history archive, and the conformed layer beneath everything else.
• Build the derived data layer: asset state classification, expected consumption and baselines, deviation scoring, an event register and governed metric definitions.
• Evaluate Things Board Trends Analytics against building our own and make the recommendation on evidence - then implement it.
• Apply proper engineering discipline: version control, testing, reproducibility. Nothing should depend on a script on someone's laptop.
• Build with large language models as a core part of the role, not an experiment: generated report narratives, plain-language explanation of anomalies, and a conversational interface over our data.
• Deliver that interface against the governed data layer, with customer isolation enforced properly and an evaluation suite that proves answers are correct.
• Maintain the boundary that matters: figures are computed in the data layer and narrated by the model, never generated by it.
• Industrialise the customer reporting our analyst currently produces by hand, so that it scales as the estate grows rather than consuming more people.
• Publish analytics outputs to our customer portal through served APIs.
  Strong Python and SQL, with production experience building and operating data pipelines - ideally in Microsoft Fabric, or a comparable lakehouse platform such as Databricks, Synapse or a dbt-based stack.
• Hands‑on experience with time‑series, sensor or telemetry data at scale, including the realities of gaps, drift, flatlines and misbehaving devices.
• Applied anomaly detection or predictive modelling taken into production, with the judgement to keep methods robust and explainable rather than elaborate.
• Practical experience building with LLM APIs beyond chat use: retrieval, tool use, generated reporting, and evaluation of model outputs.
• Self-directed and adaptable, comfortable owning delivery end to end in a small team without an engineering department to fall back on.
• Able to…