DevOps Engineer

DevOps / AIOps Engineer

Location: Remote (US) | Type: Contract (6 months) | Experience: 5+ years | Pay: $100-$115/hr

About the Role

You'll own the cloud infrastructure, deployment pipelines, and observability behind a set of AI products that are moving out of pilot and into daily use. The platform runs across Azure and AWS, with growing workload on Google Cloud for model hosting and orchestration. The first six months are weighted toward AIOps and LLMOps: giving the team real visibility into what their AI systems cost, how fast they respond, and when output quality starts to slip.

The client is a national finance and accounting advisory firm, private equity backed, that grew fast on the strength of its consulting bench and is now building its own AI products instead of buying them off the shelf. The AI group is small and early, which means the infrastructure patterns you set will be the ones the team lives with. There's no legacy platform to inherit and no committee to route around.

If you've spent the past year watching LLM features ship with no cost ceiling, no latency budget, and no way to tell when the answers got worse, this is the role where you get to fix that.

What You'll Do

• Build and maintain infrastructure as code across Azure, AWS, and GCP, primarily Terraform, with Bicep, ARM, or CloudFormation where a platform calls for it
• Own CI/CD end to end: automated testing, deployment, rollback, smoke tests, secret rotation
• Run containerized workloads on Kubernetes (AKS, EKS) and managed container services
• Stand up observability covering metrics, logs, traces, dashboards, and alerting, then push it into LLM territory: token usage, cost per call, latency, output quality drift
• Support model and agent deployment across Azure AI Foundry, AWS Bedrock, and Vertex AI
• Implement identity, secrets, and access management, including Entra ID, RBAC, Key Vault, and AWS Secrets Manager
• Keep secure end-user cloud connectivity working, including the ZTNA client deployment
• Watch the money. Cloud spend and inference costs both need visibility, and someone willing to go find the waste
• Partner with the AI and ML engineers to get RAG pipelines, multi-agent systems, and MCP-based tool integrations production ready
• Take part in incident response, on-call rotations, and release operations

Qualifications

Cloud and Platform

• 5+ years in DevOps, platform, or SRE work, with real production time in both Azure and AWS across multi-account or multi-subscription architecture, networking, and identity
• Terraform depth is required. Bicep, ARM templates, or CloudFormation is a bonus
• Kubernetes in production (AKS and/or EKS), plus Docker and Helm
• GitHub Actions or comparable GitOps tooling
• An observability stack you have actually operated rather than just installed: Prometheus and Grafana, OpenTelemetry, or equivalent
• RBAC/ABAC, SSO federation through Entra ID, and secrets management in practice
• Python for automation, and comfort working alongside AI development agents such as Claude Code or Codex
• Experience with cloud cost visibility and showback tooling, whether OpenCost or the native cloud tools
• Some history configuring secure cloud access for end users through a ZTNA product

AIOps and LLMOps

• Hands-on with at least one major AI platform: Azure AI Foundry or Azure AI Agent Service, AWS Bedrock, or GCP Vertex AI
• Exposure to LLM orchestration frameworks such as LangChain, LangGraph, or managed agent services
• Working understanding of MCP and agentic tool-use patterns
• Monitoring LLM applications for cost, latency, and quality, through LangSmith, Fiddler, or telemetry you built yourself
• RAG deployment experience: vector search, hybrid retrieval, embeddings infrastructure
• Familiarity with putting an API gateway in front of multi-model traffic for routing and rate limiting

Nice to Have

• Data pipeline tooling that feeds AI or analytics workloads
• Background supporting regulated or audit-heavy environments, SOC 2 and similar
• Azure, AWS, or GCP architect or DevOps certifications

Why Join Us

This is a small AI group inside a firm with serious enterprise reach, which is an unusual combination. You get the pace and autonomy of an early team, and the problems that come with real clients, real compliance expectations, and real money moving through the systems you support. Decisions get made in conversations, not in six-week planning cycles.

The team wants someone who has operated production infrastructure long enough to have opinions, and who is genuinely interested in what changes when the workload is a model instead of a service. Six months is enough time to build something that outlasts the contract.