DevOps / AgentOps Engineer
Revelation Pharma · United States · 3 wk ago
RemoteRemoteEngineeringFull-time
Job Activities and Responsibilities
- Design, build, and maintain CI/CD pipelines for application code, FHIR data layer, and AI agent deployments across dev/staging/production environments
- Own infrastructure-as-code (Terraform or AWS CDK) for all AWS services including HealthLake, Bedrock, Clean Rooms, Lambda, Step Functions, S3, and IAM
- Build and operate the observability stack — logging, metrics, tracing, and alerting — with particular attention to agent execution observability (token usage, latency, state transitions, error rates)
- Implement and maintain HIPAA-compliant infrastructure patterns: encryption at rest and in transit, key rotation, audit log pipelines, BAA-covered service configurations
- Work directly alongside KMS engineers during the build phase — reviewing their IaC, ensuring deployment standards are met, and absorbing system knowledge for long-term internal ownership
- Manage shared services infrastructure on the Revelation Pharma side, including cross-environment networking, secrets management, and service mesh configuration
- Automate agent deployment, versioning, rollback, and A/B testing workflows for Bedrock-based agents
- Establish and enforce environment parity, secrets hygiene, and least-privilege IAM policies across all workstreams
Required Technical Skills (Staff-Level Depth Expected)
- AWS Infrastructure (Expert): Deep hands-on experience with IAM policy design, VPC networking, KMS encryption, CloudWatch/CloudTrail, and at least 3 of: HealthLake, Bedrock, Lambda, Step Functions, ECS/Fargate, S3, Clean Rooms.
- Infrastructure-as-Code (Expert): Production experience with Terraform or AWS CDK. Must have built and maintained IaC for regulated environments — drift detection, state management, module composition, and policy-as-code (OPA/Sentinel or equivalent).
- CI/CD & Deployment (Expert): End-to-end pipeline design (GitHub Actions, CodePipeline, or equivalent). Blue/green and canary deployment patterns. Container orchestration (ECS Fargate preferred). Artifact management and promotion workflows across environments.
- Observability & Monitoring (Strong): Production experience building observability stacks — structured logging (CloudWatch Logs Insights or ELK), distributed tracing (X-Ray or OpenTelemetry), custom metrics and dashboards, PagerDuty/Opsgenie integration. Must understand healthcare-specific audit logging requirements.
- Security & Compliance (Strong): HIPAA technical safeguard implementation experience. Encryption key management (KMS), secrets management (Secrets Manager/SSM Parameter Store), network segmentation, and WAF/Shield configuration. Familiarity with HITRUST or SOC 2 control frameworks is a plus.
- Agent/AI Operations (Developing to Strong): Experience deploying and monitoring LLM-based systems in production. Understanding of prompt versioning, model endpoint management, token budget monitoring, and inference cost optimization. Experience with Bedrock specifically is preferred but not required — equivalent experience with other LLM deployment platforms (Azure OpenAI, Vertex AI, SageMaker) is acceptable.
Required Experience
- 5+ years in DevOps, SRE, or platform engineering roles with at least 2 years at a staff or senior level
- Production infrastructure experience in healthcare, fintech, or another regulated industry with formal compliance requirements
- Experience working in blended team models (internal + outsourced) — must be comfortable establishing standards that external teams follow
- Demonstrated ability to work autonomously in early-stage environments where you are defining process, not inheriting it
Preferred Qualifications
- AWS certifications (Solutions Architect Professional, DevOps Engineer Professional, or Security Specialty)
- Experience with FHIR server infrastructure (HealthLake, HAPI FHIR, or Smile CDR)
- Familiarity with Bedrock AgentCore, agent orchestration patterns, or multi-agent system deployment
- Experience with cost optimization for AI/ML workloads (inference endpoint sizing, spot instances, reserved capacity)