AI Engineer

About Agent Mercer

Autonomous commerce starts here.

A merchant should be able to describe a store and have it exist — landing page, product pages, cart, checkout — then have the money layer underneath it run without a team watching it. Routing, tokenization, subscriptions, disputes. Agents do the work. The operator signs off on anything that moves money.

Payments, tokenization, checkout, subscriptions and dispute prevention give our agents the context and tools to do the work.

The Role

Mercer works with merchants today. Orders, disputes, logistics, support and payment risk provide the context for its next action.

You will own the reasoning layer: how an agent decides, which context it can access, when it must ask for approval, and how its actions can be reviewed.

Responsibilities

• Design and ship agent workflows that read live commerce state and take real action across checkout, payments, disputes, and support
• Build the tool-use layer — typed, permissioned actions with hard limits on anything that moves money
• Own the approval boundary: which decisions an agent executes alone, which require an operator signature, and how that line is enforced in code
• Build retrieval and context assembly over orders, subscriptions, processor events, and support history
• Define evaluation harnesses and regression suites so agent behavior is measurable, not anecdotal
• Instrument every agent decision for audit — inputs, reasoning, action taken, outcome
• Work directly with the founders on which parts of a merchant’s operation should be automated next

Required Qualifications

• Four or more years of professional software engineering experience, with production systems you can point to
• Strong backend engineering in TypeScript or Python, and comfort designing APIs and data models
• Hands-on experience shipping LLM-backed features to real users — tool calling, structured output, retrieval
• Sound judgment about failure modes: hallucination, prompt injection, silent degradation, and how to contain each
• Experience building evaluation and monitoring for non-deterministic systems
• Ability to operate independently in an early-stage environment and own a domain end to end

Preferred Qualifications

• Experience with agent frameworks or multi-step orchestration in production
• Background in fintech, payments, or another domain where an incorrect action has financial consequence
• Experience with human-in-the-loop review systems or approval workflows
• Familiarity with the Claude API and its tool-use patterns

Our stack

Agents

Claude · tool use · structured output

Backend

NestJS · Node.js · TypeScript

Data

Supabase (PostgreSQL) · pgvector

Commerce

Shopify Admin & Storefront APIs (GraphQL)

Payments

Stripe · Adyen · Airwallex · PayPal

Tooling

Sentry · Linear · GitHub