Artificial Intelligence Engineer

Prospect 33; Canada; On-site; CAD 100,000 - 150,000; Full time

Job summary

A specialist AI consultancy in Canada is seeking a developer with extensive production Python experience to create innovative AI solutions in financial services. You will build key analytics systems and AI-driven pipelines, ensuring a tangible impact for major banks. The ideal candidate has a strong background in LLMs, deployed AI architectures, and cloud platforms. Join a fast-paced team that values builders who thrive under tight deadlines, offering a significant growth trajectory with real AI challenges.

Qualifications

• 5+ years shipping production Python, with service deployment experience.
• Hands-on experience with LLMs in production environments.
• Experience with API and microservices engineering.
• Strong understanding of cloud platforms like AWS, Azure, or GCP.

Responsibilities

• Build and deploy deal analytics agents.
• Create end-to-end ingestion and comprehension pipelines for unstructured banking data.
• Optimize model performance through prompt engineering.
• Ensure code is production-ready under aggressive timelines.

Skills

Production Python programming

LLM expertise

Cloud platform fluency

API and microservices engineering

Agentic AI implementation

Tools

AWS

Azure

GCP

FastAPI

Flask

Job description

One of the world’s largest banks is scaling AI across its investment banking division, with senior executive sponsorship and aggressive delivery targets. They’re building an AI engineering team inside the Client & Banking group and they need builders — not theorists, not consultants, not data scientists who dabble in code. They need someone who writes production Python every day and ships autonomous AI systems that bankers depend on to close deals, manage client relationships, and process the vast volume of unstructured documents that drive capital markets.

You’ll join through Prospect 33, a specialist AI consultancy embedded in wholesale capital markets, with a growth path that extends well beyond a single engagement. This isn’t a fire-and-forget placement — it’s the entry point to a career building AI for the most complex industry in the world.

WHAT YOU’LL BUILD

1. Deal Analytics Agents

Multi-step agentic workflows that ingest deal documents — term sheets, credit agreements, offering memoranda — extract critical terms, assess risk indicators, cross-reference against historical deals, and generate actionable summaries for senior bankers. These agents use tool-calling via MCP to query internal databases, retrieve relevant precedents, and produce structured output that feeds directly into deal review processes.

LLM-powered analytics pipelines that synthesise client data across lending, trading, and advisory products to deliver personalised insights, identify churn risk, and generate tailored recommendations. You’ll build RAG systems that ground every output in real client data, not hallucinated generalities.

End-to-end ingestion and comprehension pipelines for unstructured banking data — contracts, legal opinions, financial reports, client correspondence. Multi-format parsing, OCR, chunking strategies for long-form financial documents, embedding generation, and semantic search, all deployed as production microservices integrated with the enterprise data ecosystem.

Optimise model performance through prompt engineering, fine-tuning, and retrieval-augmented generation. Build evaluation frameworks that measure accuracy, detect hallucinations, and ensure outputs meet the deterministic confidence thresholds that regulated banking environments demand. Design modular, reusable agent components that the broader team can adopt.

MUST-HAVE REQUIREMENTS

This is a developer role that requires AI expertise — not a data science role that requires some coding. You will be writing production code every day under aggressive timelines.

• 5+ years shipping production Python — deployed services with error handling, tests, CI/CD, and monitoring. Not scripts. Not notebooks.
• 2+ years hands‑on with LLMs in production — using orchestration frameworks (LangChain, LlamaIndex, Semantic Kernel, or custom-built). You design the orchestration layer, manage context windows, handle failures, and optimise for cost and latency.
• Built and deployed RAG systems — chunking strategies, embedding models, vector stores, hybrid retrieval, reranking, and evaluation metrics (MRR, recall@k). You can defend your architecture choices.
• Agentic AI in production — autonomous agents that run without human intervention: multi‑step reasoning, tool‑calling via MCP or function‑calling, scheduled execution, self‑correction loops, and structured output that feeds downstream systems.
• Cloud platform fluency — AWS (SageMaker, Lambda, S3, Bedrock), Azure (AI Foundry, OpenAI, Cognitive Services), or GCP (Vertex AI, BigQuery, Cloud Run). Deployed and operated, not just prototyped.
• API and microservices engineering — FastAPI, Flask, or equivalent. You build clean REST APIs that other systems consume.
• Can pass a live coding assessment — Python data structures, JSON processing, algorithm implementation, system design. No open‑book.

STRONG PREFERENCE

• Financial services or regulated industry experience — capital markets, investment banking, insurance, healthcare, or government. You understand what it means to build AI where outputs must be auditable and compliant.
• Document intelligence — OCR, multi‑format parsing, processing contracts or legal documents at scale.
• Workflow orchestration — Airflow, Kubeflow, Step Functions, or similar for multi‑step AI pipelines.
• LLM evaluation and safety — hallucination detection, toxicity filtering, confidence scoring, guardrails.
• Fine‑tuning experience — LoRA, QLoRA, RLHF, or domain‑specific model adaptation with proprietary datasets.
• Prompt engineering as a discipline — few‑shot, chain‑of‑thought, structured output schemas, systematic evaluation.
• Agent observability and cost engineering — token tracking, cost attribution, guardrail monitoring, and the ability to quantify AI ROI vs. human equivalent effort.

CORE STACK

WHAT TO EXPECT

This team ships fast — weekly sprint cycles, not fortnightly. You’ll be onsite in Toronto 4–5 days a week, embedded with the banking technology group. The hiring manager is a former bulge‑bracket engineering leader who will test your coding ability in the interview. If your strength is architecture diagrams and strategy decks rather than writing code under pressure, this isn’t the right fit. If you’re a builder who thrives on velocity, this is the opportunity.

WHY PROSPECT 33

Prospect 33 is a specialist AI consultancy embedded in wholesale capital markets. Our clients are the world’s largest banks and asset managers. This isn’t a fire‑and‑forget contract placement — you’ll have a team behind you, a growth path ahead of you, and access to some of the most complex AI challenges in financial services. The problems are real, the data is massive, and the impact is measurable.