AI ENGINEER

AI ENGINEER Remote CST/EST Responsibilities: • Design and build agentic workflows: Create AI agents that can break down complex tasks, plan and reason through problems, use tools, manage context, and collaborate with other agents when needed. • Build the tools agents use: Develop well-designed functions, APIs, MCP servers, and other tools that are reliable, secure, and easy for AI agents to use effectively. • Build retrieval and knowledge pipelines: Design how agents find and use enterprise knowledge, including chunking, embeddings, hybrid search, reranking, vector and graph databases, and keeping information fresh and traceable. • Design memory and context: Build solutions that manage conversation history, long-term knowledge, session state, and audit trails while making effective use of the model's context window. • Build and improve AI evaluations: Create test datasets, regression tests, and evaluation frameworks to measure accuracy, retrieval quality, hallucinations, other failure modes, and continuously improve agent performance. • Make AI behavior observable: Instrument agents so we can understand what they're doing in production through traces, tool calls, decision paths, and other telemetry, not just traditional application errors. • Build responsible AI guardrails: Design safeguards, fallback paths, escalation processes, and human approval steps for actions where accuracy, security, or business impact really matters. • Improve performance and cost: Find the right balance between model capability, speed, and cost through model selection, routing, caching, prompt optimization, and thoughtful token usage. • Connect AI to the enterprise: Integrate agents with identity platforms, loyalty systems, CRM, CDP, data platforms, and third-party APIs to turn AI capabilities into real business solutions. • Work closely with product and business teams: Understand business problems, identify where AI can add real value, define success measures and acceptance criteria, and recognize when a simpler, deterministic solution is better than an AI agent. • Help shape our AI engineering practices: Contribute reusable patterns, standards, frameworks, and documentation that make it easier for other engineers to build secure, reliable, and scalable AI solutions. Qualifications: • 3-5+ yrs. professional software engineering experience, including code owned in production • Strong Python, with the ability to write testable, maintainable service code rather than scripts • Hands-on experience building on LLM APIs: function and tool calling, structured outputs, prompt design, streaming and error handling against non-deterministic output • Practical experience with at least one agent orchestration framework - LangGraph, LangChain, AutoGen, CrewAI, Semantic Kernel, a provider-native agent SDK or an in-house equivalent • Working experience with Retrieval-Augmented Generation: embeddings, vector stores and the retrieval quality problems that come with them • Demonstrated experience evaluating AI system quality - able to describe an evaluation suite you built, what it caught and how you acted on it • Experience with cloud providers Azure, AWS or GCP; containerization and orchestration with Docker and Kubernetes • Excellent end to end knowledge of web, client-server, services, API and database technologies • Sound judgment on when autonomy is appropriate, and the discipline to keep engineering rigor through fast iteration cycles • Excellent communication skills • Bachelor's degree in math, computer science or related tech field Preferred Qualifications: • Multi-agent systems in production, including the failure modes: sub-agent timeouts, state drift across tool calls, guardrail propagation across agent boundaries • MCP server development and agent-to-agent protocols • Observability tooling for LLM systems (LangSmith, Langfuse, Phoenix/Arize, Braintrust, Helicone or equivalent) • Prompt-injection defense and familiarity with the OWASP LLM Top 10; sandboxed code execution • Knowledge graphs, hybrid search or advanced reranking • Fine-tuning or instruction tuning (SFT/DPO/RFT), with a clear view of when it beats prompting or retrieval Data engineering depth (Databricks, Spark, S