Senior Staff AI Application Engineer

Job description

Job Description Summary We are looking for a senior full-stack engineer who has shipped production applications, and who has owned software that other engineering teams depended on — a framework, SDK, shared library or internal developer platform — living with an abstraction decision long enough to have had to correct it while consumers depended on it. Establishing engineering standards through automation rather than documentation should be instinctive. You will be the senior technical owner of GE HealthCare's internal AI engineering platform: the shared libraries, standardised interfaces, code generation tools and automated verification our data scientists and engineers use to take AI solutions into production. It exists to deliver four outcomes for the company: • reduce time from working prototype to production from months to days; • allow data scientists to ship production-grade AI without expertise in cloud infrastructure, deployment, application security or front-end development; • make security, compliance and consistency automatically enforced properties of every solution, rather than outcomes dependent on scarce expert review; • allow underlying technologies to be replaced — model provider, agent framework, user interface approach — without rewriting the solutions built on them. Your accountability is the platform's architecture and public interfaces, the automated verification that enforces our engineering standards, and the versioning and migration approach that keeps it dependable as it evolves beneath a growing number of applications. The role also includes building AI applications end to end yourself — front end, back end and agents — where that is the fastest way to prove a platform capability or unblock a priority use case; the emphasis at this level sits firmly on the platform, with application work serving it rather than competing with it. This is a hands-on senior engineering role with substantial architectural responsibility and organisational influence — not a management position, and not a research position. Job Description GE HealthCare's Chief Data and Analytics Office is delivering a portfolio of strategic AI programmes across Commercial, Finance, Supply Chain, Quality, Manufacturing and Operational Excellence. The problem this team addresses is structural. Most people creating AI solutions here are data scientists, and it would be neither reasonable nor efficient to require deep infrastructure, security and front-end expertise of them. Every solution must nonetheless reach production as a secure, reliable, observable, supportable enterprise system, and remain so through years of change. That is the central design constraint: the platform and its automated safeguards must substitute for a depth of software engineering review that is not available at the scale and pace we require. The failures a strong platform engineer prevents — undetected interface drift, poorly chosen abstractions that become permanent, over-broad access permissions, configuration that cannot be reviewed or reversed — are silent, and present months later as an outage, an audit finding or a security incident rather than as a broken feature. Core Responsibilities Architecture and design Define the platform's architecture and public interfaces: what is offered, what is guaranteed, what is private, and what a version number promises to dependent applications. Design isolation boundaries so each significant third-party technology — agent framework, model provider, message queue, data source, UI rendering approach — is confined to a single replaceable component, with automated checks making that a verified property rather than a documented intention. Validate each in practice, for example by running the same agent unmodified on two frameworks (LangGraph and AWS Bedrock AgentCore, say) with equivalent evaluated behaviour. Design the interfaces between roles and components as machine-readable specifications with automated verification, so no two teams need to read each other's code to work together reliably. Hold the boundary between declarative configuration and executable code — configuration states what , code implements how — preventing the familiar progression by which a configuration format acquires conditional logic and becomes an undocumented programming language with no type checking, debugger or tooling. Record architectural decisions formally (options rejected, trade-offs accepted, revisit conditions), and decline unnecessary abstraction in writing and with reasoning. A significant part of this role's value lies in shared capability deliberately not built. Engineering delivery and long-term ownership Lead the highest-leverage components hands-on: the agent execution layer, the secure data resolution layer, and the configuration-driven user interface system. This is a building role, not a supervisory one. Build selected AI applications end to end — front end, back end and agents — where that is the fastest way to prove a platform capability or unblock a priority use case, and judge what should then be harvested into the platform and what should remain specific to that use case. Establish automated verification before the practice it enforces is adopted — a standard that is not automatically enforced does not, in practice, exist — including automated quality evaluation of AI behaviour in the release path, so degraded output prevents release and a change of model or framework can be shown to have preserved behaviour. Own backward compatibility and the migration path. The platform will change continuously while dozens of solutions depend on it; making that change inexpensive and safe for consumers is a primary responsibility. Prevent version divergence across applications and shared components; run deprecation properly — announce, migrate, remove — and maintain visibility of the platform's own engineering health, acting before it is raised. Interfaces with adjacent disciplines Define and maintain the interfaces between this team and adjacent specialist functions — infrastructure, security, data — as version-controlled specifications with automated validation rather than request queues or standing meetings. These cover what an environment provides, what an infrastructure component accepts, what a service requires to run and is permitted to do, what must pass before release, what artefact is deployed and how it is promoted, and what telemetry and service level commitments apply. Anticipate and act on the recognised ways such boundaries deteriorate: a request queue re-forming, resources created outside the sanctioned path, release throughput constrained by a single function, divergent standards emerging, or incidents beginning with a dispute about ownership. Represent the platform's architecture and trade-offs to enterprise architecture, information security, risk and senior leadership, in terms appropriate to each audience. Technical leadership Mentor engineers and data scientists moving into platform and application engineering, and establish a review culture in which human review addresses design judgement and business fit, because formatting, type correctness, security fundamentals and interface compliance are handled automatically. Author the guidance, reference implementations and worked examples that let colleagues and AI coding assistants work productively without direct supervision, and set the standard for responsible use of those assistants under named human accountability. Experience Requirements Bachelor's degree in Computer Science, Software Engineering or a related field, or equivalent demonstrable practical experience. An advanced degree is welcome but not required. Minimum 8 years of professional software engineering experience, including demonstrable ownership of a shared library, framework, SDK or internal developer platform relied upon by other engineering teams. This is the essential requirement: not a record of delivering many applications, but an engineer whose software other engineers' software depends on, and who can describe concretely what it cost when an abstraction had to change. Software engineering Expert Python: advanced static and structural typing (mypy or Pyright, strict), interface definition, and judgement about where type-level guarantees earn their cost. Strong TypeScript and modern component-based front-end engineering (React), sufficient to design typed component contracts and a design token system rather than only consume them. API and interface design and long-term evolution: specification (OpenAPI), generated clients, versioning, compatibility, deprecation. Relational databases and SQL at depth: schema design, migrations, indexing, behaviour under load, injection resistance established structurally rather than by review. Testing strategy across unit, integration, interface-contract and end-to-end levels; CI/CD design (GitHub Actions, GitLab CI) including shared reusable pipeline components, and the discipline of keeping checks fast enough that engineers do not bypass them. AI and Generative AI engineering Substantial production experience with LLM systems: instruction design, tool and function calling, structured output, retrieval-augmented generation and retrieval quality, cost and latency control, characteristic failure modes. Hands-on experience with more than one AI agent framework (for example LangGraph, LangChain, AWS Bedrock AgentCore, OpenAI Agents SDK) and demonstrated ability to design a durable abstraction across them — including rigorous reasoning about which capabilities belong in a common, framework-independent agent definition and which do not. Familiarity with emerging standards for AI tool and context interoperability (for example the Model Context Protocol), with the judgement to design toward such a standard's shape without prematurely building infrastructure for it. Evaluation-driven development at depth: designing evaluation suites and quality measures for non-deterministic systems and using them as a formal release gate. Cloud and delivery engineering Infrastructure and production operations are specialist disciplines owned elsewhere in the organisation; our applications request what they need through version-controlled declarations validated automatically, rather than by authoring infrastructure or access policies directly. You are not expected to author infrastructure modules, access policies or network components, nor to own cloud estate design, release execution or infrastructure on call. Required at this depth: Strong working knowledge of a major public cloud, ideally AWS: serverless compute (Lambda), messaging (SQS), object and vector storage (S3, S3 Vectors), secret management, managed relational databases (Aurora), content delivery, event routing, managed AI model services (Bedrock) — sufficient to design well against them and reason about cost, quota, scaling and failure. Infrastructure-as-code: fluent reading and review (Terraform) — interpret an execution plan, evaluate whether a component's interface meets a requirement, articulate precisely what is missing and why. Access-management design literacy at boundary level: what a grant permits, what a boundary constrains, why least-privilege access is better generated than hand-written, and how to identify an over-broad request in review. Observability design (OpenTelemetry): distributed tracing, structured logging, standardised telemetry, service level objectives and error budgets. Supply chain and delivery security: immutable artefact identity, build provenance, secret scanning, dependency vulnerability management, short-lived federated credentials. Cost engineering at the level of declared resource requirements: compute sizing, execution limits, model consumption, caching and result limits. Leadership and communication Ability to explain a technical design and its trade-offs to executive leadership, to security and risk, and to a data scientist, adjusting the level without losing the substance — and willingness to state clearly when a programme is blocked and what would unblock it. Experience leading work where the principal difficulty was organisational rather than technical, with a concrete account of how a cross-team boundary was kept healthy. Additional Information Relocation Assistance Provided: Yes