Lead AI Engineer || Toronto / Minneapolis, Minnesota (Hybrid)

Lead AI Engineer

Toronto / Minneapolis, Minnesota (Hybrid)

Phone Video

Job Description:

Design, build, and own production agentic AI systems end-to-end across the Wealth Management platform.

This is the core builder role on a greenfield AI squad. You will design and ship production-grade agentic AI systems — multi-tool reasoning agents, RAG pipelines, tool integrations, and LLM orchestration layers — solving real enterprise problems across financial services workflows. You won't just prototype; you own systems through deployment, monitoring, and iteration. The team is new, patterns are yours to define, and the problems are high-value with direct business impact. This role may grow into a technical lead position as the team scales.

WHAT YOU'LL BUILD

• Multi-tool ReAct agentic systems with LLM-driven reasoning loops, tool chaining, and state management
• RAG pipelines — ingestion, chunking strategy, hybrid retrieval (vector keyword), freshness management, citation grounding
• Custom tool integration layers connecting AI agents to enterprise systems and internal APIs
• Streaming backends, session management, rate limiting, and enterprise-grade hardening
• Eval frameworks — golden datasets, LLM-as-judge, regression detection in CI/CD
• Tiered LLM routing for cost / latency / quality optimization

REQUIRED EXPERIENCE

• 8–10 years total engineering experience — with 2–3 years specifically building production agentic or LLM systems (not just prototypes) and 5–6 years in software engineering, backend systems, or adjacent technical roles
• Hands-on RAG architecture — chunking tradeoffs, retrieval failures, evaluation
• Built or extended tool integration layers connecting LLM agents to external systems
• Strong Python backend — FastAPI, async, Pydantic, streaming responses
• Deployed on Kubernetes / OpenShift with Vault, health probes, CI/CD
• Can diagnose a RAG system returning wrong answers — not just "reprompt it"
• Knows when NOT to use agents — cost/complexity tradeoff thinking
• NICE TO HAVE
• Experience with AWS Bedrock, Azure OpenAI, or enterprise LLM gateway patterns
• Experience with structured data query generation (NL-to-SQL) and output validation
• Financial services or regulated industry background
• LoRA / PEFT fine-tuning with clear reasoning on when to fine-tune vs. prompt engineer
• Prior consulting delivery — understands shipping under real constraints

TECH STACK

Python · FastAPI · Pydantic

LangChain · LangGraph

ReAct · Plan-and-Execute · Agentic Orchestration

RAG · Hybrid Vector Search

Function Calling · Tool Use

Multi-Agent Systems

Structured Outputs · JSON Mode

SSE Streaming · WebSockets

Async Python · Concurrency

Kubernetes · Container Orchestration

Enterprise LLM Platforms

Secrets Management

CI/CD Pipelines

PostgreSQL · Redis · Vector DB

Observability · Tracing · LangSmith

Guardrails · PII Detection

Fine-tuning · PEFT / LoRA