Role Overview
Build production AI systems for clients. You'll design and ship agentic workflows on Claude and AWS Bedrock AgentCore, retrieval systems backed by Bedrock Knowledge Bases, and MCP-based automations that connect directly into client tool stacks (Notion, Linear, HubSpot, AWS, Google Workspace, and more). This is a hands-on build role. You will write code, deploy infrastructure, and sit in client rooms explaining why the system behaves the way it does.
Tasks & Responsibilities
- Design and build applied AI systems on Claude and AWS Bedrock AgentCore: RAG pipelines backed by Bedrock Knowledge Bases, agentic workflows, and multi-tool orchestration via MCP.
- Own delivery of AI components end-to-end: architecture, implementation, evaluation, and post-launch monitoring.
- Integrate Claude-based agents with client infrastructure and business tools via MCP servers (HubSpot, Notion, Linear, Google Workspace).
- Translate ambiguous client problems into scoped technical builds with clear success criteria and numbers attached.
- Document architecture and decisions in the client's Notion record so infra details never live only in someone's head.
- Work alongside Pre-Sales Engineers during scoping calls to validate technical feasibility before commitments are made.
- Select and tune the right model, retrieval strategy, and Knowledge Base configuration per use case.
- Build and maintain eval suites (offline test sets, regression checks, LLM-as-judge scoring) that catch quality drift before a client does.
- Design agentic control flow: tool-selection logic, retry/fallback paths, and guardrails against runaway loops or unsafe tool calls.
- Instrument production systems with logging, tracing, and cost/latency dashboards to debug a bad output from a client.
- Write modular infrastructure as code (Terraform) for AI workloads on AWS.
- Conduct architecture reviews on other engineers' AI builds and flag failure modes before they ship.
Competencies
- Can explain to a client, in plain language, why an agent made the tool call it made
- Comfortable debugging a production LLM failure under time pressure, not just in a notebook
- Can wire an MCP server into an agent and debug why a tool call didn't fire
Education, Experience & Language
- Master's degree in Computer Science or equivalent
- Previous experience building and shipping production ML/AI systems
- Fluent in English and Arabic
Tools & Certifications
- Strong Python engineering; comfort with AWS
- Practical experience with LLM application patterns: RAG, agentic tool-use via MCP, prompt engineering, evaluation design
- Hands-on experience with AWS Bedrock AgentCore and Bedrock Knowledge Bases
- Nice to have: AWS Certified Machine Learning Engineer – Associate, AWS Certified Generative AI Developer – Professional, AWS Certified Solutions Architect – Associate or Professional, AWS Certified Developer – Associate, AWS Certified Security – Specialty (relevant given our compliance posture)
