Senior AI Engineer

Role - Senior AI Engineer

Location - Toronto ,ON

Mandatory Skills:

• JVM Engineering (Java/Kotlin)

• Knowledge Graphs (Neo4j, Cypher, GraphRAG)

• MCP (Model Context Protocol)

• Production Platform Engineering (CI/CD, Docker, Kubernetes, Observability)

• Document Generation & Rendering (Apache POI, PDFBox, pptxgenjs, Office Open XML, Word/PPT/PDF generation, template management)

We are looking for an experienced and highly motivated Senior AI Engineer to join Canada segment’s Technology Strategy team.

In this role, you will design and build the agentic workflow that breaks a complex analytical engagement into stages, coordinates large language models and tools, and keeps every result grounded in source evidence. You will be accountable for how the system stores and retrieves knowledge, how it exposes its capabilities to other agents and assistants, and how it proves that what it produces is faithful to the source.

Position Responsibilities

You are an expert in agentic systems, knowledge graphs, AI-enabled document generation and rendering, with extensive knowledge in cloud infrastructure, the AI service stack, CI/CD, and observability. You will work closely with engineering and business partners to deliver an AI system that is accurate, explainable, safe to operate, and presentable to the business.

1. Design and build the agentic analysis workflow

• Model a long-running, multi-stage analytical workflow as a sequence of AI-agent steps that can pause, resume, and recover without losing work.

• Orchestrate large language models and tools (retrieval, calculation, rules, and structured extraction) into reliable, traceable steps.

• Engineer prompts and structured-output contracts so each step returns validated, machine-usable results rather than free text.

• Build on an event-sourced service framework (for example, the Akka SDK) where state, decisions, and evidence are persisted at every step.

2. Build and operate the knowledge graph

• Design and run the knowledge graph that stores entities, relationships, and provenance extracted from the source corpus (for example, Neo4j or MongoDB Atlas).

• Build the ingestion and extraction pipeline that turns raw documents into a validated, queryable graph, with every fact citing its source.

• Implement graph-based and vector retrieval (GraphRAG and embeddings) so agents answer from grounded evidence rather than memory.

3. Expose capabilities through open agent interfaces

• Build Model Context Protocol (MCP) tool servers and agent-to-agent (A2A) interfaces, so the system's capabilities are available to external assistants and to larger multi-agent workflows.

• Integrate with MCP-compatible clients and low-code agent builders, for example Microsoft Copilot Studio agents, so business users can query the graph and trigger analysis conversationally.

• Build a conversational interface that answers questions strictly from the grounded knowledge graph.

4. Set a high bar for grounding, evaluation, and safety

• Implement anti-hallucination controls: every output statement traces to a cited source, and anything unsupported is flagged or withheld.

• Build automated evaluation (section-level scoring, regression tests, edge-case coverage, and human-review rubrics) and turn it into clear acceptance criteria.

• Add guardrails for prompt injection, data minimization, and safe output, with explicit human-in-the-loop checkpoints, and produce the evidence that AI and model-governance review will ask for.

Preferred Background

Experience delivering enterprise-scale AI systems in production environments.

Experience integrating AI capabilities into business workflows and customer-facing experiences.

Strong collaboration skills with both technical and business stakeholders.

Passion for innovation, continuous improvement, and responsible AI practices.