The Principal AI Operations Engineer reports to the Sr. Director, AI Center of Excellence, in the Technology Department. This role defines how AI systems at Inspira Financial are instrumented, measured, monitored, and evidenced, and drives adoption of those standards across the enterprise, spanning both vendor-provided and internally built AI capabilities. Primary duties include building the enterprise telemetry and observability standard for AI systems; designing evaluation, drift detection, and agent guardrail frameworks; translating AI risk and compliance frameworks into operational controls and audit evidence; maintaining the enterprise AI system inventory; and teaching other engineering teams to adopt these standards independently.
Duties & Responsibilities:- Define the enterprise telemetry standard for AI systems, covering traces, prompts, tool calls, retrieval, tokens, latency, and cost, built on a vendor-neutral schema such as OpenTelemetry GenAI semantic conventions.
- Build the logging and observability layer that provides the AI Center of Excellence, business units, and the AI Governance Committee with a single view of AI activity across custom applications, agent platforms, and vendor-embedded AI.
- Integrate AI telemetry with enterprise log aggregation, application performance monitoring, security monitoring, alerting, and incident response platforms to support operational visibility, investigation, and audit retention.
- Establish retention, redaction, masking, and access patterns for AI conversation and trace data that satisfy privacy, records retention, and legal hold obligations.
- Publish reference implementations and starter kits so teams adopting the telemetry standard inherit a working path.
- Design the evaluation framework for AI systems, including golden datasets, offline evaluation harnesses, automated and LLM-as-judge scoring, and regression gates in CI.
- Define drift criteria for each class of AI system and build detection for output quality degradation, retrieval decay, input distribution shift, and silent model updates from providers or vendors.
- Build runtime guardrails and boundary enforcement for agents, including tool and action authorization, scope and spend limits, escalation paths, human-in-the-loop triggers, and hard-stop conditions.
- Establish the AI incident model, including severity taxonomy, detection, alerting, rollback, postmortem process, and conversion of incidents into new test cases.
- Translate NIST AI RMF, ISO/IEC 42001, model risk management practice, and applicable privacy and recordkeeping obligations into operational controls engineers can implement.
- Build the evidence layer that automates capture of artifacts needed for internal audit, external examination, and client due diligence.
- Integrate AI operational data with governance, risk, and compliance platforms to support continuous control monitoring, automated evidence collection, risk reporting, and audit readiness.
- Maintain the operational AI inventory across custom builds, agents, integrations, and vendor AI features, including an owner, risk classification, and control coverage for each entry.
- Define the assurance pattern for vendor AI, including required telemetry, attestations, detection of vendor model changes, and compensating monitoring.
- Build metric frameworks connecting AI usage, quality, cost, and risk posture to business outcomes, and report these to CoE leadership, business unit leaders, and the AI Governance Committee.
- Own token and platform cost visibility and attribution by team and use case and lead resulting optimization work.
- Run working sessions, document patterns, review designs, and coach teams through their first fully instrumented AI deployment.
- Partner with Risk, Compliance, Legal, Privacy, and Security to incorporate feedback from the risk triad and ensure adequate detection, response, and mitigation measures are in place.
- Perform other duties as assigned.
Education & Experience:Minimum required:
- Bachelor of Science in Engineering or Computer Science
- Relevant certifications are a plus: Datadog Certified Professional, Google Cloud Professional Data Engineer for BigQuery ML, or PagerDuty's AI Ops
- 8 or more years of telemetry-driven engineering experience, including 2 years operating LLM, ML, or agentic systems in production, with hands-on ownership of instrumentation and monitoring.
Skills & Abilities:- Production-grade proficiency in Python and SQL, including building tooling, pipelines, and services other teams depend on.
- Demonstrated experience in AI evaluation and observability, including eval harnesses, tracing, scoring, drift detection, and guardrails.
- Working command of at least one AI risk or governance framework, such as NIST AI RMF, ISO/IEC 42001, EU AI Act, or model risk management practice, and the ability to convert it into controls and code.
- Experience operating in a regulated environment where logging, retention, evidence, and access control are enforced rather than aspirational.
- Ability to influence without formal authority across engineering teams, legal counsel, and executive stakeholders.
- Financial services, healthcare, benefits administration, or other regulated industry experience preferred, particularly environments handling PHI, PII, or fiduciary data.
- Depth of knowledge with telemetry platforms and related metric and measurement practices preferred.
- Hands-on experience integrating and operating enterprise platforms for AI and application observability, centralized logging, governance, risk, continuous compliance monitoring, AI system inventory, evaluation and testing, incident management, and cloud cost management preferred.
- FinOps or cloud cost management experience applied to token and inference spend preferred.
- AI security fluency preferred, including prompt injection defense, OWASP LLM and agentic risk categories, red teaming, tool sandboxing, and least-privilege design for agent identities.
- Track record of standing up a capability that other teams adopted voluntarily preferred.
Other Requirements:- Ability to occasionally work outside normal business hours to meet project deadlines as needed.
- Prolonged periods of sitting at a desk and working on a computer.
