AI and Agentic Engineer II
Exelixis
August 07, 2026
On-site
Alameda, CA
IT
Summary/Job Purpose
Engineer II - AI and Agentic designs, builds, and deploys production-grade agentic applications that automate multi-step IT and business workflows.
Responsibilities
- Design and deliver end-to-end agentic applications and multi-agent workflows to automate IT and business processes.
- Convert prototypes into production-ready services with clear specs, modular design, robust error handling, and test coverage.
- Integrate LLMs, prompts, tools, APIs, and enterprise data sources (e.g., AWS Bedrock, Databricks Mosaic AI, MCP servers).
- Implement CI/CD, automated evaluation harnesses, observability, and artifact versioning for AI/agent workloads.
- Write architecture notes, runbooks, SOPs, and technical documentation.
- Lead prompt/spec reviews, code reviews, threat modeling, and release-readiness gates.
- Apply secure-by-design and responsible AI principles (data governance, privacy, safe prompting).
- Contribute reusable components, patterns, and internal libraries.
- Partner with product, security, and platform teams to deliver scalable solutions.
Education/Experience
- Bachelor’s (5 years) or Master’s (3 years) in CS/AI/Data Science/Software Engineering (or equivalent).
- 2–4 years building AI applications/automation, including LLM-enabled apps, agentic workflows, and API integrations.
Required Skills
- Intermediate–Advanced Python and/or JavaScript/TypeScript (modular design, testing, packaging).
- Agent frameworks/patterns (tool use, planning/execution loops, guardrails, evaluation, prompt/version management).
- REST APIs; OAuth/JWT, secrets management, secure data handling.
- CI/CD, automated unit/integration testing, linting, observability.
- Security best practices, architecture design principles, responsible AI/data governance.
- Communication/collaboration; mentoring junior engineers.
- Cloud AI platforms (prefer AWS/Bedsrock, Databricks Mosaic AI) and vector stores/RAG.
- AI-native coding agents (Claude Code/Codex/Copilot or equivalent).
Preferred
- Evaluation frameworks, red-teaming, or LLM observability tooling.
Engineer II - AI and Agentic designs, builds, and deploys production-grade agentic applications that automate multi-step IT and business workflows.
Responsibilities
- Design and deliver end-to-end agentic applications and multi-agent workflows to automate IT and business processes.
- Convert prototypes into production-ready services with clear specs, modular design, robust error handling, and test coverage.
- Integrate LLMs, prompts, tools, APIs, and enterprise data sources (e.g., AWS Bedrock, Databricks Mosaic AI, MCP servers).
- Implement CI/CD, automated evaluation harnesses, observability, and artifact versioning for AI/agent workloads.
- Write architecture notes, runbooks, SOPs, and technical documentation.
- Lead prompt/spec reviews, code reviews, threat modeling, and release-readiness gates.
- Apply secure-by-design and responsible AI principles (data governance, privacy, safe prompting).
- Contribute reusable components, patterns, and internal libraries.
- Partner with product, security, and platform teams to deliver scalable solutions.
Education/Experience
- Bachelor’s (5 years) or Master’s (3 years) in CS/AI/Data Science/Software Engineering (or equivalent).
- 2–4 years building AI applications/automation, including LLM-enabled apps, agentic workflows, and API integrations.
Required Skills
- Intermediate–Advanced Python and/or JavaScript/TypeScript (modular design, testing, packaging).
- Agent frameworks/patterns (tool use, planning/execution loops, guardrails, evaluation, prompt/version management).
- REST APIs; OAuth/JWT, secrets management, secure data handling.
- CI/CD, automated unit/integration testing, linting, observability.
- Security best practices, architecture design principles, responsible AI/data governance.
- Communication/collaboration; mentoring junior engineers.
- Cloud AI platforms (prefer AWS/Bedsrock, Databricks Mosaic AI) and vector stores/RAG.
- AI-native coding agents (Claude Code/Codex/Copilot or equivalent).
Preferred
- Evaluation frameworks, red-teaming, or LLM observability tooling.