Job Description & Details
This is a senior-level technical anchor role where you will embed directly with client teams to ship governed, production-grade agentic systems built on Anthropic's Claude stack. Rather than just building flashy demos, you will own the end-to-end architecture and implementation of AI agents that actually survive the harsh realities of enterprise production. It is a high-impact position that sits at the intersection of hands-on engineering, solution architecture, and client leadership.
What You'll Actually Be Doing
You will spend your days bridging the gap between business outcomes and complex technical execution, deploying agentic workflows using frameworks like Bounteous Arc and Claude Code. You'll lead integrated squads of engineers, build custom Model Context Protocol (MCP) servers to connect Claude to messy enterprise data, and establish rigorous evaluation suites to gate every release. Expect to navigate tough architectural trade-offs, balancing autonomy against risk, and working closely with client security and risk stakeholders to ensure agents operate safely in regulated environments.
The Core Tech Stack
You must possess deep, hands-on fluency with the Claude API, Claude Agent SDK, and MCP server development, alongside daily proficiency with Claude Code for agentic workflows. Python and TypeScript are your primary languages, backed by solid cloud deployment skills, preferably on AWS. The role also demands advanced prompt and context engineering techniques—such as progressive disclosure, compaction, and subagent context isolation—because enterprise data is messy and your agents need to stay accurate across multi-step tasks without blowing up token budgets.
Interview Expectations
Expect the hiring team to present a complex scenario involving an autonomous agent hallucinating or looping indefinitely on a live enterprise database, and ask how you would debug and constrain it using prompt caching and human-in-the-loop checkpoints. They will also likely test your architectural pragmatism by asking you to defend when to build a simple prompt-chaining workflow versus when to introduce a complex orchestrator-subagent design. The interviewer is secretly looking for your ability to balance engineering elegance with cost predictability and real-world risk management.
Application Advice
Your resume needs to immediately prove you have shipped LLM agents to actual production environments, not just built local Jupyter notebooks or PoCs. Make sure to prominently feature keywords like Claude API, MCP servers, Python, TypeScript, and AWS architecture. Weave these technologies into narrative bullet points that highlight how you managed token costs, built evaluation frameworks, and solved complex data integration challenges for enterprise clients.