AI Solutions Developer

AI Solutions Developer - (GenAI / OpenAI) — Financial Crimes Technology (FCT)

Location: Montreal, QC

Description du poste

We are looking for a hands-on AI Solutions Developer (backend) to help us build AI-enabled capabilities on top of large language models (LLMs) - starting with automated generation of client / customer summaries (e.g., relationship summaries, risk summaries, KYC/CDD profile narratives, and alert/case briefings) from the structured data we already hold. You will integrate OpenAI (and similar) APIs into our backend services, assemble the data and context these models need, design effective prompts, and ship reliable, well-governed features that summarize complex client information into clear, accurate, audit-ready text. This is a backend / integration role suited to a developer who is comfortable pulling data from multiple sources, calling LLM APIs, shaping the output, and exposing it through services - and who cares about accuracy, safety, and cost in a regulated financial-services environment. We are open to either Java or Python as the primary language.

Responsabilités clés

• Build AI-enabled features that generate client summaries from structured and unstructured data using OpenAI APIs and similar LLM providers.
• Integrate LLM calls into application services - prompt construction, model selection, structured outputs (JSON), streaming responses, token-budget and cost management, retries, and error handling.
• Assemble context for the model - retrieve and shape data from databases and internal APIs into the structured input the model needs to produce each summary type.
• Design and iterate on prompts and templates for different summary "types," and build lightweight evaluation to check accuracy, completeness, and faithfulness (no hallucinated facts).
• Expose capabilities through clean backend APIs so other teams can consume them.
• Apply responsible-AI guardrails - PII handling/redaction, prompt-injection defenses, output validation, and audit logging suitable for a regulated environment.
• Collaborate with engineering, business analysts, and compliance stakeholders to define what a "good" summary looks like and to validate outputs.
• Use AI coding assistants (e.g., GitHub Copilot, ChatGPT) to accelerate your own development.

Qualifications requises (Must Have)

• 3-5 years of professional software development experience as a hands-on developer (junior to mid-level).
• Strong proficiency in at least one backend language - Java (with Spring Boot) and/or Python (FastAPI / Flask). We are open to either.
• Hands-on experience calling LLM APIs - particularly the OpenAI API (chat completions, function/tool calling, structured/JSON outputs, embeddings) - or a strong willingness and demonstrated aptitude to ramp up quickly.
• Practical prompt-engineering skills - writing, testing, and iterating on prompts to get reliable, well-structured output.
• Experience integrating with data sources - relational databases (SQL) and REST APIs.
• Solid understanding of RESTful API design and JSON.
• Good engineering fundamentals - version control (Git), testing, and writing maintainable code.
• Awareness of data privacy / PII concerns and a careful, quality-focused mindset.
• Strong problem-solving and communication skills; able to work with non-technical stakeholders to shape requirements.
• Bachelor's degree in Computer Science, Engineering, or related field (or equivalent experience).

Familiarité / À titre souhaitable (tout sous-ensemble est un atout)

• LLM / orchestration libraries - Spring AI, LangChain / LangChain4j, LlamaIndex, Semantic Kernel, or equivalent.
• Other LLM providers - Anthropic Claude, Google Gemini, or open-source models (Llama, Mistral).
• Streaming responses (SSE / WebSockets) and async patterns for LLM output.
• Evaluation tooling for LLM outputs - measuring faithfulness, hallucination, and task success; familiarity with Langfuse / LangSmith or similar observability for AI.
• Cloud experience (Azure / AWS / GCP) and containerization (Docker).
• CI/CD (Jenkins or similar) and build tooling (Gradle / Maven for Java, or pip/poetry for Python).
• Responsible-AI / model-governance practices in a regulated context.
• Exposure to financial-services, AML/KYC, compliance, or risk domains.