Agentic AI Engineer

Stack

Responsibilities

• ▹Shape technical direction and architecture: orchestration, agent frameworks, tool integrations (including MCP), memory and state management, evaluation, and observability
• ▹Write production code at least half your time, particularly during the team's first year
• ▹Stand up CI/CD, testing, evaluation, and deployment infrastructure for agentic systems
• ▹Graduate prototypes into hardened, production-grade systems under a "you build it, you run it" model
• ▹Design for enterprise from day one: multi-tenant isolation, secrets management, audit logging, role-based access controls, human-in-the-loop controls
• ▹Partner with Security Engineering on threat modeling for agentic architectures (prompt injection, tool misuse, data exfiltration)
• ▹Coach power users and departmental teams; build templates, SDKs, and sandboxes for internal builders
• ▹Partner with the Data, Analytics & Systems team on the datasets and pipelines agentic systems depend on
• ▹Set the bar for code quality, testing/evaluation, documentation, and on-call practices; drive technical hiring; mentor engineers

Requirements

• ▹7+ years of professional software engineering experience building production systems, with strong systems design fundamentals
• ▹Hands-on experience building production systems that integrate with LLMs and/or agentic patterns: orchestration, tool use, memory and state management, evaluation, observability
• ▹Demonstrated understanding of optimizing workloads across deterministic and non-deterministic capabilities, striking the right architectural balance
• ▹Production experience with at least two of: Python, TypeScript/Node.js, Go; comfort working across the stack
• ▹Hands-on expertise with LLM APIs (OpenAI, Anthropic), agentic frameworks (LangChain, CrewAI), RAG over business content (Confluence, contracts, policies), vector databases (pgvector, Pinecone), workflow automation (n8n, Langflow), and LLM observability/evaluation tooling (LangSmith, Arize)
• ▹Track record of going from zero to one: a platform, function, or product area built up from scratch and scaled
• ▹Experience operating in regulated or security-sensitive environments; solid grasp of enterprise security fundamentals
• ▹Comfortable exercising technical leadership independent of positional authority
• ▹Product-first approach, shipping code quickly with real-world impact in mind
• ▹Enjoy ownership and building key pieces of platforms

Soft skills

About the company

Benchling is the AI platform for biotech R&D: scientists use it to design experiments, capture structured data, and run AI agents and models in their workflows; trusted by over 200,000 scientists, including Sanofi, Moderna, and more than half of the world's top 50 biopharma companies.