Description

Responsibilities:

• Build and run production AI agents: develop TypeScript agents and services on AWS (Lambda, Step Functions, DynamoDB, Bedrock) using Strands or LangGraph and AI SDKs.
• Design agent behavior: decide what the model investigates and what plain code enforces. 
• Sandbox and constrain agents: build read-only runtimes, scoped tool access, worktree-contained reads, and budget and timeout guards that hold when the model misbehaves.
• Write evals for everything AI-driven: build golden sets and replays from real production cases, calibrate judges, and gate changes on recall, precision and false-positive rate. Run model and effort-level comparisons before a change ships.
• Make agents observable: emit reason-coded metrics and structured logs and write alarms that fire on real failures. Investigate production misses in Splunk and VictoriaMetrics.
• Run infrastructure and pipelines: maintain Terraform roots across environments under our tagging and permission-boundary rules, and improve GitLab CI/CD for build, test, deploy and quality gates.
• Handle integrations that survive retries, rate limits, permanent delivery failures and stale state.
• Use AI to write and review most code, and keep agent context current.
• Review AI-written changes rigorously: check security, correctness and maintainability, and catch what the tests don't.
• Enable other teams: set up assistant configuration, skills, hooks and MCP servers, pair with engineers, and turn what works into reusable practice.
• Build the surfaces agents need: React and API work for review summaries, dashboards and internal tooling, preferring output that agents, CLIs and Slack can read.
• Document how it works: sequence and architecture diagrams, runbooks, and short decision records that explain why.

Requirements:

• LLM features in production: has shipped features on LLM APIs (prompts, tool calling, structured output, long context) and can describe the eval that gated release, what failed and what changed.
• TypeScript: strong production TypeScript and Node, including async code, testing and monorepo tooling.
• Python: working proficiency for evals, analysis and scripting.
• AWS serverless: Lambda, Step Functions, DynamoDB, IAM and event-driven design, plus Bedrock.
• Infrastructure as code: able to read and change Terraform safely across environments.
• Data: hands-on experience with NoSQL and SQL stores, including schema changes and migrations on live systems.
• Observability: able to design metrics, logs and alarms for something non-deterministic, and to debug it from them.
• CI/CD: proficiency with CI/CD tools, ideally GitLab CI/CD.
• AI-assisted development: daily use of an AI coding assistant, with the discipline to verify what it produces.
• Communication: fluent written and spoken English, and the ability to explain a trade-off and get a team to adopt an approach.

Nice to have:

• Multi-agent systems, MCP servers or evaluation frameworks (for example Strands Evals).
• LLM cost control: model routing, prompt caching, token budgets.
• Prompt-injection defense and agent security.
• React and GraphQL for internal tools.
• Java and Spring Boot, for reading or integrating with existing services.
• Code intelligence or retrieval work (indexing, symbol search, knowledge bundles).
• Kubernetes, AWS AppSync and EventBridge.
• Playwright/Cucumber test automation.
• Claude Certified Developer or Claude Certified Architect (Anthropic)

Key Skills

Education

Any Graduate