Claude Platform Engineering

Works alongside delivery teams on early Claude engagements, explaining model tiers, context windows, knowledge cutoffs and token pricing when asked, and rough-costing a simple workload. Structures basic prompts with clear instructions, XML tags and examples, makes a first Messages API call with the SDK, and reads back roles, content blocks, stop reasons and usage. Handles keys and rate limits carefully, and talks accurately about tool-use cycles, RAG, evals, usage-policy limits, core LLM risks and the API/Bedrock/Vertex access paths.

Builds and demos small Claude features under supervision. Picks a model tier and justifies it, shows Projects, artifacts and connectors credibly, and notes feature gaps across API, Bedrock and Vertex. Writes system prompts that hold role, constraints and output format, applies chain-of-thought or extended thinking, prefill and stop sequences, and refines prompts against captured failures. Implements multi-turn state, streaming, vision and file inputs, retries, schema-defined tools and tool loops, a basic RAG pipeline with citations, small eval sets, privacy and guardrail basics, and provisions Claude on Bedrock and Vertex.

Owns production Claude workloads end to end. Designs model selection and fallback, migrates versions with eval evidence, and sets client expectations on real capability boundaries. Engineers long-context and parameterized prompts, curates context across turns, uses caching, grounding and structured outputs, and works to token, latency and cost budgets with telemetry in place. Builds composable, least-privilege tool sets with resilient error handling, tunes and measures retrieval quality, runs CI regression and red-team evals, adds injection defenses and human-in-the-loop controls, and designs multi-cloud routing and failover.

Acts as the technical authority across multiple engagements. Reviews solution designs, prompts, integrations, tool definitions, knowledge architectures and multi-cloud deployments for cost, security, reliability and compliance fit, and hardens them before production. Chooses between RAG, long context, MCP resources and knowledge tools per use case, establishes prompt versioning and tool design standards, builds shared client libraries and eval infrastructure, gates releases on eval results, supports client security reviews with Trust Center artifacts, and keeps teams current on model and platform changes.

Shapes how the whole practice builds on Claude. Sets platform standards, model adoption and upgrade strategy, and multi-cloud platform direction across a client portfolio. Defines prompt and context engineering standards with reusable libraries, API architecture standards, enterprise knowledge-architecture patterns and tool ecosystems spanning a client's application estate. Owns the evaluation methodology and quality bars, responsible-AI standards and incident escalation paths, and advises client leadership on AI quality measurement and acceptance criteria.