AI Automation Engineer

Greenlight Consulting•Toronto, ON

5h•CA$110,000 - CA$140,000•Hybrid

About The Position

Greenlight helps organizations solve complex business challenges through intelligent automation, agentic AI, and custom technology solutions. Our teams work directly with clients to understand their operations, identify opportunities, and rapidly build solutions that create measurable business value. We combine deep consulting expertise with hands-on engineering to bridge the gap between strategy and execution. We’re building a future where consultants and engineers work alongside AI to deliver faster outcomes, stronger businesses, and transformative customer experiences. We use Anthropic’s Claude as a core delivery tool — embedded in how we design, build, and validate automation solutions — and this role works inside that environment every day. The AI Automation Engineer is the execution layer of the Greenlight delivery pod. You take a signed-off Solution Design Document and build it — cleanly, efficiently, and to a standard that holds up in production. This is not a back-office development role. You work in a lean, fast-moving pod alongside an Automation Business Consultant who owns the process logic and AI specifications, and an AI Delivery Engagement Manager who owns the client. Your job is to build what the spec says, flag what the spec missed, and close the gap between design and working software in two-week sprints. As AI agent-based automation grows alongside our RPA practice, this role requires genuine fluency across both — UiPath for process automation, and Claude-powered agent workflows for AI-driven solutions. The developer who thrives here is comfortable in both worlds and curious about where they intersect.

Requirements

3–7 years of hands-on development experience, with at least 2 years on UiPath a client-facing or professional services environment
Demonstrated experience building and deploying AI agent or LLM-integrated solutions — not just familiarity; you have shipped something that uses an LLM in production
Strong understanding of automation architecture: reusability patterns, exception handling frameworks, Orchestrator asset management, and queue-based processing
Experience working within agile delivery cycles — sprint-based development, backlog grooming, iterative releases
Direct exposure to enterprise system integrations — REST APIs, ERPs, CRMs, document platforms — and the real-world messiness that comes with them
Precision builder — your code is clean, documented, and maintainable; you do not leave technical debt for the next person to clean up
AI-native — you understand how LLMs behave, where they fail, and how to build against probabilistic outputs reliably
Problem-solver, not ticket-closer — when the spec is incomplete or the integration is messier than expected, you figure it out and document what you found
Fast and focused — two-week sprints are real; you scope your effort, hit your commits, and flag early when something is bigger than it looked
UiPath Studio, Orchestrator, REFramework — proficient across the full development and deployment lifecycle
Anthropic Claude API or OpenAI API — tool use, structured outputs, prompt chaining, output validation
REST API integration — authentication patterns, error handling, payload management
JSON, Python, or C# — enough to build custom activities, handle data transformations, and extend UiPath where needed
Postman or equivalent API testing tools for pre-integration validation
Jira or Azure DevOps for sprint tracking and defect management

Nice To Haves

Microsoft Power Platform (Power Automate, Power Apps) awareness is a plus

Responsibilities

Build automation solutions in UiPath Studio — efficient, well-structured, maintainable, and easy to understand; scalability and reusability are not optional
Develop and configure AI agent workflows using Anthropic’s Claude API — integrating LLM components, defining tool use, and wiring agent logic into the broader automation pipeline
Validate the AI-generated SDD before build begins — flag gaps, ambiguities, or technical constraints early rather than discovering them mid-sprint
Create reusable components, activity libraries, assets, and queues via UiPath Orchestrator and Studio that raise the quality floor for every subsequent build
Integrate automations with client systems — ERPs, CRMs, document management platforms, APIs — handling authentication, error handling, and edge cases the spec may not have fully anticipated
Write and execute unit tests and pre-UAT test procedures — own quality before the client ever sees it
Participate in UAT alongside the Automation Business Consultant — triaging defects, distinguishing genuine build issues from scope queries, and resolving assigned bugs with speed and quality
Build and configure agents: define tool schemas, implement decision logic, handle multi-turn interactions, and ensure output formats match the finalized specifications
Implement prompt-to-output pipelines — taking the prompt architecture designed and wiring it into the automation workflow reliably and testably
Handle LLM failure modes in code: output validation, fallback logic, human-in-the-loop triggers, and graceful degradation when model outputs fall outside acceptable parameters
Work with JSON editors, API testing tools, and the Anthropic API directly to validate agent behaviour before it reaches a client environment
Write Python, JavaScript, or C# beyond the UiPath activity layer — custom scripts, data transformation pipelines, lightweight utilities, and integration logic that the platform can’t handle natively
Build and maintain custom UiPath activities and libraries where out-of-the-box components fall short — extending the platform rather than working around it
Design and implement API integrations from scratch where no connector exists — authentication, payload handling, retry logic, error management, and response parsing
Develop lightweight back-end components to support agent workflows — data preprocessing, output formatting, webhook handlers, and queue management that sit outside the automation itself
Work within two-week sprint cycles — sprint planning, daily standups, reviews, and retrospectives; you are an active participant, not a resource that takes tickets
Respond quickly to build-phase queries from the Automation Business Consultant — ambiguity in the spec costs sprint time; resolve it fast and document the outcome
Flag technical risks, integration blockers, and scope gaps to the AI Delivery Engagement Manager as they surface — not after they have already caused a delay
Support go-live and post-deployment stabilisation — the build is not done until the client is running it confidently in production