AI Observability- Integration & Enablement

**LOCATION: Remote Type: Contract**

About Newpage Solutions

Newpage Solutions is a global digital health innovation company helping people live longer, healthier lives. We partner with life sciences organisations which include, pharmaceutical, biotech and healthcare leaders, to build transformative AI and data driven technologies addressing real-world health challenges.

From strategy and research to UX design and agile development, we deliver and validate impactful solutions using lean, human-centered practices.

We are proud to be a ‘Great Place to Work®’ certified company for the last three consecutive years. We also hold a top Glassdoor rating and are named among the “Top 50 Most Promising Healthcare Solution Providers” by CIOReview. As an organisation, we foster creativity, continuous learning and inclusivity, creating an environment where bold ideas thrive and make a measurable difference in people’s lives.

Your Mission

The organisation runs a shared observability platform for LLM-based applications. This role connects that platform to the systems around it, so that teams get observability as a property of the environment they already work in rather than as something they have to go and set up. Two examples of what that means in practice. When a team is provisioned with LLM credentials or deploys an agent, observability should follow automatically: the project, the credentials, the collection configuration and the ownership metadata all created without anyone raising a request. Separately, where results need to appear somewhere other than the observability tooling, such as an internal agent registry that records each agent alongside its evaluation scores, this role builds and owns that path. This is systems integration engineering. It is not building or operating the collection pipeline and observability backends, which sit with the platform engineering roles, and it is not writing instrumentation inside application teams’ codebases. It is the connective tissue between the platform and everything else, and the level of service end users experience depends almost entirely on how well it works..

What You’ll Do

Make observability automatic.

• Integrate with stack provisioning so that granting LLM credentials, model access or deploying an agent creates the corresponding observability without a separate request.
• Wire identity and access through, so project access follows existing team membership and group structure rather than being granted by hand.
• Seed sensible defaults at creation: collection configuration, ownership metadata, retention, and redaction settings appropriate to the workload.
• Handle the whole lifecycle rather than just creation, including ownership changes, decommissioning, and keeping both systems consistent when they drift apart.

Integrate with the systems that consume the data.

• Publish evaluation scores into the internal agent registry, so that a registered agent carries its current results wherever it is looked up.
• Deliver alerts and signals into the tools teams already work in, rather than requiring them to watch another place.
• Agree the data contract for each integration and version it, so that downstream consumers do not break when something changes upstream.

Keep the integrations dependable.

• Build for partial failure: retries, idempotency, backfill, and reconciliation when two systems disagree about the same thing.
• Monitor the integrations themselves, so that a broken feed is noticed here rather than reported by a user.
• Document what each integration does and what it guarantees, so other teams can depend on it without having to ask.
• Turn recurring manual requests into automation, so that the service improves without the team growing

What You Bring

• Backend integration engineering in Python, TypeScript or both. APIs, webhooks, event-driven messaging, queues and authentication flows, to a standard that stands up in production.
• Designing and consuming APIs between internal systems, including agreeing data contracts and evolving a schema without breaking consumers.
• Identity and access integration. OIDC or SSO, service accounts, API key lifecycle, and mapping existing groups to roles and permissions.
• Building for failure. Idempotency, retry and backoff, partial-failure handling, and reconciling state between systems that disagree.
• CI/CD and infrastructure-as-code literacy, enough to hook cleanly into provisioning pi…