AI Engineer

Kobie Marketing•Detroit, MI

17h•Remote

About The Position

Kobie is seeking a hands-on AI Engineer to join their team and contribute to building an internal agent platform. This platform automates analyst workflows, surfaces insights from program data, and provides an LLM-native way to interact with complex loyalty logic. The role involves building agent harnesses, writing tools for these agents, and ensuring the reliability and evaluation of production features. This is a practical, shipping-focused role rather than a research position, requiring individuals who can reason carefully, write working code, and adapt to new tools independently. Kobie values demonstrated impact and a track record of building robust systems.

Requirements

3+ years of professional Python experience, including production experience building and operating services.
1+ years of hands-on work with LLMs in production, covering prompt/context engineering, tool/function calling, structured outputs, and RAG.
Working knowledge of LangChain/LangGraph or a comparable framework (e.g., AgentCore Strands, CrewAI, Semantic Kernel).
Experience with LLM observability tools (e.g., Amazon CloudWatch, LangSmith, Langfuse, MLflow, OpenTelemetry).
Experience designing evaluation frameworks (e.g., MLFlow, DeepEval, LLM-as-judge, multi-turn regression).
Fluency with Git, Docker, and modern API frameworks.
Clear written communication skills and the judgment to determine when a feature is ready for production.
Equivalent practical experience (bootcamps, self-taught work, career changes, non-CS technical degrees) is accepted in lieu of a bachelor's degree.

Nice To Haves

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, and understanding of semantic models.
Familiarity with multi-agent patterns (supervisor/router, subagent/handoff, reflection, human-in-the-loop).
A considered perspective on agent application and the ability to push back when an agent is not the appropriate solution.
Experience in Loyalty, MarTech, AdTech, or a comparable data-rich B2B domain.

Responsibilities

Build agent harnesses in Python using LangChain and LangGraph, incorporating features like tool-calling, structured outputs (Pydantic/JSON schema), retries, streaming, and memory.
Package agent harnesses for the AgentCore Runtime, including necessary context, tools, skills, and subagents for production flows.
Write the tools and skills that agents utilize, such as API integrations, SQL queries against Snowflake, and Snowflake-backed knowledge retrieval with Pydantic validation.
Build evaluation harnesses (golden datasets, LLM-as-judge, regression suites) using AgentCore Evaluations and integrate them into CI pipelines.
Implement guardrails around tool execution, including auth scoping, input/output validation, PII and prompt-injection protections, and hallucination mitigation.
Prototype, deploy through Amazon AgentCore, monitor traces, and resolve issues for shipped features.
Collaborate with data engineers on Snowflake-backed retrieval patterns.
Contribute to refining internal engineering patterns as the technology stack evolves.