AI Engineer

Kobie is a recognized leader in loyalty solutions, known for its innovative and collaborative culture. They are seeking a hands-on AI Engineer to build and maintain their internal agent platform, focusing on automating workflows and enhancing reliability in production environments.

Responsibilities:

• Build agent harnesses in Python using LangChain and LangGraph, including tool-calling, structured outputs (Pydantic/JSON schema), retries, streaming, and memory
• Package agent harnesses for the AgentCore Runtime with appropriate context, tools, skills, and subagents that fit cleanly into production flows and scenarios
• Write the tools and skills agents use API integrations, SQL queries against Snowflake, Snowflake backed knowledge retrieval with clear contracts and Pydantic validation
• Build evaluation harnesses (golden datasets, LLM-as-judge, regression suites) using AgentCore Evaluations, and wire them into CI
• Implement guardrails around tool execution: auth scoping, input/output validation, PII and prompt-injection protections, and hallucination mitigation
• You own what you ship: prototype, deploy through Amazon AgentCore, monitor traces, and fix it when it breaks
• Partner with data engineers on Snowflake backed retrieval patterns (Cortex Analyst and Cortex Search Services)
• Contribute to refining our internal engineering patterns as the stack evolves

Requirements:

• 3+ years of professional Python, with production experience building and operating services
• 1+ years of hands-on work with LLMs in production: prompt/context engineering, tool/function calling, structured outputs, RAG
• Working knowledge of LangChain/LangGraph or a comparable framework like AgentCore Strands, CrewAI, or Semantic Kernel
• Experience with LLM observability tools: Amazon CloudWatch, LangSmith, Langfuse, MLflow, or OpenTelemetry
• Experience designing evaluation frameworks (MLFlow, DeepEval, LLM-as-judge, multi-turn regression)
• Fluency with Git, Docker, and modern API frameworks
• Clear written communication and the judgment to know when something is ready to ship
• A bachelor's degree is not required. Equivalent practical experience: including bootcamps, self-taught work, career changes, or non-CS technical degrees counts
• Hands-on experience with Amazon Bedrock and/or AgentCore as a developer: runtime, gateways, memory, policy, guardrails, observability, awscli, evaluations
• Experience with Snowflake, Snowpark, or Snowflake Cortex
• Fluency in writing and reading SQL, as well as understanding semantic models
• Familiarity with multi-agent patterns: supervisor/router, subagent/handoff, reflection, human-in-the-loop
• A considered view on where agents should and shouldn't act and comfort pushing back when 'let's add an agent' isn't the right answer
• Experience in Loyalty, MarTech, AdTech, or a comparable data rich B2B domain