Founding AI Engineer

Kouvre; Montreal (administrative region); Hybrid; CAD 120,000 - 180,000; Full time

Job summary

Kouvre, the AI agent for restaurants, is looking for its Founding AI Engineer in Montreal. You will own the brain of the system, building the reasoning layer, action agents, and evals that make our platform feel like a sharp employee who never drops the ball.

Partner with the Founding CTO to turn plain-text messages into concrete actions—draft promos, publish campaigns, respond in the restaurant voice, and continuously improve reliability, judgment, and operational edge across multiple delivery

Qualifications

• Experience building real LLM systems with tool use and orchestration.
• Ability to ship from notebook to production with measurable impact.
• Comfort handling ambiguous user goals and fuzzy signals.
• Strong reasoning about latency, cost and trade-offs.

Responsibilities

• Design the reasoning layer mapping text to actions across platforms.
• Build action agents to draft promos, campaigns, and social posts.
• Develop an eval pipeline to validate improvements.
• Define safeguards and judgment guardrails.
• Collaborate with CTO to align product with business goals.

Skills

LLM systems

Tool use

Orchestration

Eval pipelines

Failure modes

Tools

Python

LangChain

LLM tooling

Job description

All roles

Founding AI Engineer

Full-time·Montreal (Hybrid)

Kouvre is the AI agent for restaurants. Independent owners text it like an employee and it runs the business: promos, menus, prices, reviews, their website, all across UberEats, DoorDash, and Skip at once. No dashboards. We're already making restaurants money. One BOGO at a Montreal spot, launched by a single text, lifted UberEats revenue by 1,000% on a slow Tuesday.

As our Founding AI Engineer, you own the brain. You will partner closely with the Founding CTO to turn a plain-text message into the right action, building the reasoning, the agents, and the evals that make Kouvre feel less like software and more like a sharp employee who never drops the ball.

The Mission

• For the owner: Make Kouvre understand intent. When an owner texts "fill my slow Tuesday," Kouvre has to know what that means and do the right thing: the promo, the platform, the timing, without a brief.
• For the company: Make the agent trustworthy. The gap between a fun demo and a business owners depend on is everything you'll build: reliability, judgment, and knowing what not to do.

What You'll Build

• The reasoning layer: The agent logic that maps a casual text to the correct, complete action across platforms, handling ambiguity the way a good employee would.
• The action agents: Generative agents that draft promo copy, build a campaign, reply to a review in the restaurant's voice, or turn a new dish into a social post.
• The eval pipeline: The harness that proves a change made the agent better, not just different. Your safeguard against shipping regressions into someone's livelihood.
• The judgment guardrails: Knowing when to act, when to confirm, and when to stay quiet. A 10/10 action or nothing.
• The continuous edge: Feedback loops that make the agent sharper with every restaurant it runs.

What We're Looking For

• Deep agent intuition: You've built real LLM systems (tool-use, orchestration, retrieval, evals) and you know their failure modes cold.
• Product instinct: You care about the owner's outcome, not model benchmarks. You think "did it do the right thing," not "was the output plausible."
• Ship-it capability: You take an idea from notebook to production and own the path to real users.
• Scrappy with ambiguity: Small datasets, messy real-world signals, and fuzzy goals energize you.
• Clear thinking: You can explain a precision, latency, and cost trade-off plainly and make the call.

Why This is Different

• AI is the product. You're building the thing the company is, not a feature on a dashboard.
• Founding equity, and the rare chance to define an AI discipline from scratch, with a CTO who backs you.
• Live proof it works. An operator-led company that's already making restaurants money.
• Zero bureaucracy. Proof of work over process. The best agent is one an owner forgets is software.
• Montreal-based and hybrid. Work from home, and meet in person when it matters.