Tech Lead - Agent Development

You know that moment when your coding agent passes all its tests, coverage is at 100%, and something feels off? You dig in, and you realize it modified the assertions instead of fixing the code. It cheated. If you're the kind of person who has already been through that, who figured out why it happens and who architected the solution so it doesn't happen again, we have something for you.

At Vooban, we build agentic systems for a major partner in a highly regulated sector. We're talking about complete delivery chains: from business specification all the way to code deployed in production, with agents at every link. Coding agents that consume structured specs and generate code, tests and documentation. Design agents that produce traceable architecture diagrams. Analysis agents that turn meeting transcripts into formal requirements. All in an environment where governance and security are non-negotiable.

The Tech Lead role is about seeing the whole system. Not just your agent, but the chain. Identifying where the bottleneck moves: today it's the code, tomorrow it's the spec, the day after it's the missing knowledge the agent needs to make the right architecture decision. You standardize practices across teams that don't have the same maturity. You coach three audiences in parallel:

• People whose job isn't code, business analysts, architects, designers, so they can work with agents in GitHub Copilot, Claude Code and VS Code.
• Senior developers who know more than you about their stack but need to change the way they work.
• And of course, your own team, which you help grow technically every day. It's as much change management as it is technical.

If you're at the stage where you've built systems with LLMs in production and you're looking for the place where your impact will multiply across a team, this is it.

Note: the role requires a regular presence at our Quebec City or Montreal offices, with availability at the client's site.

More concretely, you will be called upon to

• Design end-to-end agentic architectures: orchestrators, builders, reviewers, with automated validation loops (quality, security, business compliance). Decide when an agent needs autonomy versus when a sequential chain is enough.
• Industrialize the SDLC through agentic AI. The goal: compress the specification-to-production cycle from weeks to hours. That means orchestrating fleets of coding agents, not just one human driving one agent.
• Build and structure the knowledge bases that allow agents to make architecture decisions consistent with what already exists: internal patterns, in-house libraries, service contracts, organizational standards.
• Coach mixed teams (developers, analysts, architects) to adopt agentic workflows. The challenge isn't technical, it's changing how experienced people work.
• Standardize practices across teams of different maturity levels: repo structure, spec templates, agent sharing, harmonization of instructions and skills.
• Be the go-to person when an agent system goes off the rails: exploding tokens, overly ambiguous specs that cause hallucinations, agents making undocumented architecture decisions.

Your profile

• 7 to 10 years of experience in software development, including 1 to 2 years with LLMs in production.
• Mastery of Python. Java and/or GoLang are major assets.
• Real expertise in agent orchestration (LangChain, LangGraph, CrewAI, AutoGen, Claude Code, or proprietary solutions). You know what the framework gives you and what you have to build yourself.
• Cloud experience, ideally AWS (Bedrock, Landing Zone).
• Experience with or understanding of regulated environments (finance, insurance, healthcare).
• Bachelor's degree in computer science, software engineering or a related field. A solid self-taught background with hands-on experience, we look at that too.
• Able to influence without hierarchical authority. You coach people who have more years than you, and they listen to you because you're credible.