Description
Cloud Engineer II
This Statement of Work covers a six-week Phase 0 proof-of-concept engagement. Phase 0 is intended to validate feasibility, not to deliver an enterprise-scale production service.
This position will deploy an open-weight model endpoint in the client's infrastructure in AWS, connect it to OpenCode, and provide enough evidence for the client to decide whether to proceed to Phase 1.
The Phase 0 business case is based on validating whether open-weight inference can work for non-Opus coding workloads, with success criteria including model quality, latency, throughput, end-to-end developer flow, developer satisfaction, and a go/no-go decision for the next phase.
Responsibilities
- Deploy and configure an AWS-hosted, open-weight coding model endpoint using client-provided infrastructure.
- Configure model-serving runtime, GPU allocation, endpoint exposure, health checks, service routing, basic error logging, credential handling, and repeatable deployment artifacts.
- Expose the model endpoint through an OpenAI-compatible API path capable of supporting representative coding prompts.
- Configure and validate OpenCode integration with the model endpoint, including any required adapter, proxy, gateway, or compatibility layer.
- Demonstrate end-to-end OpenCode-to-model request flow for representative coding-agent workflows.
- Demonstrate representative coding-assistance use cases, including code completion, code explanation, unit test generation, refactoring assistance, error diagnosis, small feature implementation, tool usage, repository/file-context prompts where supported prompts and agent skills, and OpenCode/Open Spec workflows.
- Collect and save basic performance data in Grafana, including time to first token, end-to-end latency, tokens per second, concurrency behavior, error rate, GPU utilization, and memory utilization.
- Implement baseline observability for the POC, including logs, health checks, basic metrics, and simple dashboard or metric views where feasible.
- Provide reusable deployment and configuration artifacts, such as Helm values, Kubernetes manifests, scripts, runtime configuration, gateway/proxy configuration, runbook, troubleshooting notes, and environment assumptions.
- Support limited developer validation with a small client-designated pilot group if access, timing, and dependencies allow.
- Support a lightweight quality comparison between the open-weight model and client's current coding-assistant experience if time and access allow.
- Provide preliminary Phase 1 sizing inputs, including throughput observations, bottlenecks, latency drivers, scaling considerations, recommended AWS instance families, and remaining unknowns.
- Deliver a written final report covering architecture, deployment, OpenCode integration results, functional test results, performance measurements, technical gaps, security and operational considerations, cost and scale observations, reusable artifacts, and Phase 1 readiness.
- Deliver a clear Phase 1 go/no-go recommendation, including required changes, unresolved risks, and recommended next steps for AWS-based enterprise-scale implementation.
- Conduct a knowledge-transfer session with stakeholders at the end of the engagement.
Requirements
Required Qualifications
3-5+ years of Cloud Engineering experience.
Hands-on AWS, Kubernetes, Docker, and API integration experience.
Experience with infrastructure automation (Terraform, Helm, Kubernetes manifests, or similar).
Strong experience with monitoring and observability tools such as Grafana and CloudWatch.
Ability to troubleshoot distributed cloud environments and performance issues.
Preferred Qualifications
Experience with Generative AI, LLMs, or AI infrastructure.
Familiarity with model-serving platforms such as vLLM, Triton, TGI, or similar.
Experience supporting GPU-based workloads in AWS.
Knowledge of OpenAI-compatible APIs and AI-powered developer tools.