Applied AI Engineer (Agents & Systems Integrations)

**Applied AI Engineer (Agents \& Systems Integrations)**

*SunMed Growers / Tidal Creek Growers*

**About the role**

We’re building an AI\-native operating layer for the business — not another chatbot experiment. As our Applied AI Engineer (Agents \& Systems Integrations), you’ll design, ship, and run the agents and integrations that connect real workflows to our systems of record, using Model Context Protocol (MCP) servers, Claude, AWS, and business APIs.

Think virtual teammates: durable agents that take work, call tools, update SoRs, and hand results back to people — then get better as more processes come online.

This isn’t a traditional software\-engineer seat (no endless ticket queue, no feature factory). You’ll own the build\-once patterns — MCPs, APIs, AWS\-hosted agents/apps, Claude\-powered workflows — and the practical craft of making AI teammates trustworthy enough that the org actually uses them. You bring the engineering fundamentals; we collaborate with you on AI design and build on top of them.

You’ll work day to day with IT and the people who own each business process, and you’ll be regularly onsite at SMG/TCG. If you want to build agentic systems that run a real company, not a demo, this is that seat.

**What you’ll own**

· Design, build, and operate AI agents that execute real business workflows

· MCP / tool / API integrations into systems of record and internal services

· AWS\-hosted agents and AI\-created apps (build\-once, reuse everywhere)

· Hand\-off patterns so people can assign work to virtual teammates as capabilities mature

· Observability for every agent: structured logging, step\-by\-step traces of what an agent did and why, metrics, dashboards, and alerts that tell us when something breaks or drifts

· Reliability and incident response: health checks, retries and safe failure, a pause or kill switch, replaying a failed run to find the cause, and fixes backed by regression tests — production agents, not one\-off demos

· Cost and quality monitoring: tracking AI usage and spend, response speed, error rates, and output quality over time, including after model or prompt changes

· Security and governance for agents: least\-privilege access, secrets handling, audit logging, and clear rules for what data may and may not reach an outside AI model

· Partner with operators and process owners to turn messy workflows into agent\-ready systems

**Must\-haves**

· Strong software engineering fundamentals and API/integration chops: you arrive able to build, test, and ship production software on your own

· Comfortable in Python and/or TypeScript for production work

· Proven experience shipping agentic or tool\-using AI systems (agents, copilots, or automation that calls tools/APIs — not just chatting with an LLM)

· Ability to own work end\-to\-end: design → build → deploy → monitor → improve

· Observability mindset: builds logging, tracing, metrics, and alerting into agents from day one, and can diagnose a failing agent from its telemetry

· Clear written and verbal communication with technical and non\-technical partners

· Bias for practical, reusable systems over one\-off glue

· Security\-minded engineering: least\-privilege access, safe handling of credentials and secrets, audit trails, and good judgment about company data and AI tools

**Strongly preferred**

· Hands\-on with MCP (Model Context Protocol) or equivalent tool\-calling agent frameworks

· Experience integrating with business SoRs / ERPs / SaaS APIs

· AWS (or similar cloud) for hosting agents, functions, or small apps

· Claude / Anthropic (or comparable frontier LLM) production usage

· Evaluation, guardrails, and operational discipline for AI in production

· Hands\-on with observability tooling (for example OpenTelemetry, AWS CloudWatch, or an LLM tracing tool such as Langfuse) and with alerting and incident response

· Experience defining service\-level targets (uptime, error rate, response time) and monitoring cost and quality for AI workloads

**Nice\-to\-haves**

· Prior work building “virtual teammate” / multi\-agent patterns

· Data plumbing (queues, webhooks, light ETL) around agent workflows

· Familiarity with operations\-heavy industries (manufacturing, logistics, regulated environments)

· Light frontend skills for internal operator UIs

**Details**

· Full\-time W\-2

· Hybrid (not fully remote); regularly onsite at SMG/TCG, exact schedule set with HR

· Comp: $115,000–$130,000

Pay: $115,000\. 00 \- $130,000\. 00 per year

Benefits:

• 401(k) matching
• Dental insurance
• Employee assistance program
• Employee discount
• Flexible spending account
• Health insurance
• Health savings account
• Life insurance
• Paid time off
• Parental leave
• Vision insurance

Application Question(s):

• Show me code or a service you wrote in Python or TypeScript that runs in production. How is it tested and deployed?
• Tell me about connecting to a business system like an ERP or SaaS tool through its API (application programming interface). What was hard?
• Describe something you took from idea to design, build, deploy and ongoing support. What did you have to fix after launch?
• Tell me about something you built once that ended up being reused for other needs. How did you design it so that could happen?
• An agent will read and update our business systems. How do you keep it safe, and what company data should never go to an outside AI model?
• Walk me through an AI agent or tool\-using system you built that real people used. What did it do, and what went wrong?
• How do you decide which tools an agent gets, and how do you stop it from calling the wrong one or looping?
• What have you built or used with MCP, a standard way to connect AI to tools and data? How did you handle sign\-in and permissions?
• How do you know an agent is working, and how do you catch it getting worse after a change?
• What have you run in production with Claude or a comparable model, and how did you manage cost, speed and quality?
• How should a person assign work to an agent, check it, and take over when it's unsure?
• Have you built agents that work together or hand work to each other, or a long\-running agent that acts like a teammate? How did you keep them from stepping on each other?
• At 2 a.m. an agent starts producing wrong results. How do you find out, and how do you find the cause from its logs and traces?
• What logging, metrics, tracing and alerting tools have you used, and how did you decide what to measure?
• How do you make sure a change doesn't break something that already works?

Work Location

In person