Venesky-Brown’s client, a public sector organisation in Edinburgh, is currently looking to recruit an AI Python Engineer for an initial 12 month contract with option to extend on a rate of £600/day (Outside IR35). This role will be based in Edinburgh, however, attendance at the project site will only be required on an as-needed basis.
Responsibilities:
– Build and maintain the shared Python platform library that all workloads depend on — configuration, logging/telemetry, Azure clients, model interface abstractions.
– Hold a high engineering bar across the codebase: type safety, test coverage, linting, dependency hygiene.
– Keep the library’s abstractions clean as models, transports, and workloads rotate underneath them.
– Implement and maintain Temporal-based workflow workers for the document processing pipeline (ingestion → extraction → reasoning/rule-assertion → deterministic mapping).
– Build the plumbing that loads and serves open-weight models inside workers (embedded inference engine pattern), including model provenance verification and warm-load behaviour.
– Implement per-queue scaling, priority isolation, and burst handling.
– Develop and maintain Terraform for the platform estate across non-prod, pre-prod and prod environments.
– Own the GitOps deployment path (ArgoCD) and the container build/publish pipeline into the registry.
– Operate workloads on AKS — namespaces, autoscaling (KEDA), service mesh, policy and security add-ons.
– Build telemetry, dashboards and alerting (Managed Prometheus / Grafana, App Insights) shaped for first-line consumption.
– Implement support automation as a first-class platform layer — self-healing operator patterns, runbooks-as-code — to minimise manual operational handover. (Directly supports the RUN-readiness risk on this programme.)
– Implement validation and verification logic so extracted/derived data meets quality standards before it leaves the pipeline.
– Integrate the platform with enterprise systems (message bus, databases, document stores) and support the AI engineers in wiring new model workloads in.
– Work to the team’s design and task discipline (low-level design templates, tightly scoped tasks, ADO tracking).
– Document architecture, runbooks and operational guidance to support deployment and ongoing support.
Essential Skills:
– Strong, demonstrable production Python — typed code (mypy/strict or equivalent), testing (pytest), linting, packaging and dependency management.
– Containers and Kubernetes in production: building images, deploying and operating workloads, debugging in-cluster.
– Infrastructure as code — Terraform (or equivalent) with a modular, environment-driven structure.
– CI/CD and GitOps — automated build/test/deploy pipelines; declarative deployment.
– Cloud platform engineering, ideally Azure (AKS, Service Bus, managed Postgres, Key Vault, Blob Storage, managed identity).
– Observability — metrics, logs, traces; building dashboards and alerts, not just consuming them.
– Comfort working around AI/ML workloads — integrating model-serving runtimes, understanding inference resource behaviour — without needing to own model science.
– Experience delivering and operating services end-to-end, including the support and maintenance phase.
– Awareness of secure handling of sensitive data and relevant data-protection obligations.
Desirable Skills:
– Temporal.io or another durable-execution / workflow-orchestration framework.
– vLLM or similar LLM-serving runtimes; familiarity with GPU workload scheduling on Kubernetes.
– KEDA, Istio (or another service mesh), ArgoCD.
– Experience supporting A/B model rollouts behind a stable interface (the reasoning queue has a model-swap on the roadmap).
– Vector search infrastructure (e.g. Milvus / self-hosted vector DB on Kubernetes) — a candidate roadmap component.
– Experience contributing to platform support-automation / SRE-style operability as a deliverable in its own right.
– Exposure to regulated / public-sector delivery and associated governance.
If you would like to hear more about this opportunity please get in touch.
…