Role: FDE/AI Architect (W2-Requirement)
Location: Remote
Duration: Long term
Must Have Skils: Claude (skills, plugins, MCP servers/connectors), LLM-based applications
Job Description
The AI CoE Pod Lead is a hands-on builder who designs, ships and runs Claude skills, plugins and agent apps for our client’s internal business teams, while acting as the technical lead for the Pod's Forward Deployed Engineers (FDEs). Roughly 50-60% of the role is building and deploying; the rest is technical guidance, design review and stakeholder alignment.
Key responsibilities
Build and ship (hands-on)
Design, write and iterate Claude skills (SKILL.md instructions, scripts, templates, reference assets) that encode business team workflows and produce reliable, repeatable outputs.
Package skills, commands, sub-agents, hooks and MCP connectors into plugins that internal teams can install and use with minimal setup.
Build agent apps on the Claude API and Agent SDK where a workflow needs orchestration, state, scheduling or integration beyond a skill.
Build and maintain MCP servers and connectors to systems (CRM, ticketing, document stores, data platforms) with least-privilege access.
Write evaluation suites for each skill and agent (golden datasets, regression tests, output-quality rubrics) and use them to gate releases.
Deploy and operate
Own the release path for skills and plugins: version control, code review, CI/CD, environment promotion and rollback.
Manage distribution and access through the organization's plugin catalog and admin controls, so the right teams get the right capabilities.
Set up monitoring for usage, adoption, quality, latency and token cost; act on drift and failures.
Lead the technical side of workstream transitions: takeover code and context from incumbent teams, close technical debt and environment gaps, and prepare cutover plans.
Technical leadership
Act as technical lead for the Pod's FDEs: set standards for skill design, prompt and context engineering, testing and documentation.
Run design and code reviews; unblock FDEs on hard problems through pairing and reference implementations.
Maintain a shared library of reusable skill patterns, templates and connectors to speed up new builds.
Decide when a need is best met by a skill, a plugin, an agent app or a conventional integration, and document the reasoning.
Business engagement
Work with business owners to turn workflows into clear requirements, success metrics and PRDs.
Run discovery and demos; drive UAT with business users and fold feedback into fast iterations.
Train business teams on using and contributing to skills; build internal champions.
Governance, security and quality
Prepare the technical evidence for each acceptance criterion: architecture docs and ADRs, UAT results, InfoSec reviews, deployment readiness checklists.
Partner with InfoSec on data classification, access controls, PII handling and prompt-injection defenses for skills, connectors and agents.
Track and report cost, usage and value delivered per workstream.
Required qualifications
10+ years in software engineering, with 2+ years building and shipping LLM-based applications to production.
Hands-on experience building with Claude: skills, plugins, MCP servers/connectors, the Claude API, and tool use; or equivalent depth on a comparable LLM platform with a clear ramp to Claude.
Strong Python and/or TypeScript; comfortable with REST APIs, JSON schemas, auth (OAuth, service accounts) and scripting.
Working knowledge of agent patterns: tool calling, sub-agents, context management, retrieval, memory and human-in-the-loop checkpoints.
Experience designing LLM evaluations and regression testing for non-deterministic outputs.
Practical DevOps: Git workflows, CI/CD, cloud (AWS, Azure or Google Cloud Platform), secrets management, logging and monitoring.
2+ years as a tech lead or player-coach guiding engineers through design reviews and code reviews.
Proven ability to work directly with non-technical business users, translate workflows into requirements, and run UAT.
Sound grasp of AI security and data governance: PII handling, access control, prompt injection and data-exfiltration risks.
Preferred qualifications
Experience with the Claude Agent SDK, Claude Code, plugin catalogs and enterprise admin controls.
Track record of taking over in-flight AI systems from another team and stabilizing them for production.
Background in domains: EHS, ESG, operational risk, product stewardship or supply chain risk.
Experience with enterprise systems commonly connected to agents: Salesforce, ServiceNow, Microsoft 365, SharePoint, Snowflake or Databricks.
Experience building reusable internal platforms or developer tooling adopted by multiple teams.
Familiarity with AI governance frameworks (NIST AI RMF, ISO/IEC 42001) and SOC 2 controls.
Experience in a global delivery model with partner or offshore engineering teams.
