We’re building autonomous AI agents that resolve customer support tickets for e-commerce businesses end-to-end. They cancel orders, issue refunds, check inventory, talk to suppliers, and get better over time.

To make this scalable, we’re building an AI Manager: a higher-level agent that trains, monitors, and optimizes the Support Agents with minimal human input - effectively an AI-run support team.

You lean toward product, UX, and the user experience. You’ll build the interfaces and workflows that make self-serve setup actually work and talk to users to figure out what’s missing.

What you’ll do

• Ship production features across the stack
• Work in tight loops: build → test → measure → refine
• Use AI tools daily (Cursor, Claude, etc.) and push them further; automate glue work, generate scaffolds, build evals, multiply your own output
• Take ownership of outcomes, not tickets

What we’re looking for

We don’t care about years of experience. We care that you can build.
You might be a fit if:

• You can own features end-to-end and enjoy doing so
• You ship fast and care about quality, clean architecture, solid tests, good UI
• You’re resourceful and persistent, you figure things out
• You’ve built things you’re proud of (products, open source, side projects)
• You use AI tools daily and want to keep raising that bar
• You think in systems and multipliers, not just tasks

Stack

• Backend: Python, FastAPI, AWS
• Frontend: TypeScript, React, Next.js

What makes this technically hard

1. Making AI fully self-serve

Other platforms either rely on implementation teams to configure agents for each customer, or stay relatively shallow. We want a store owner or customer support manager to set up, train, and improve their agent entirely on their own.

That means building feedback and evaluation systems that let non-technical users correct agents the way they’d manage a human colleague: “you shouldn’t have refunded that,” “check inventory before responding.”
Not tuning parameters. Not writing prompts. This is an unsolved problem.

2. AI-native integration into arbitrary systems

Every customer has a different stack: Shopify, custom ERPs, legacy order systems with undocumented APIs. The typical approach—build each integration by hand—doesn’t scale.

We’re building AI that can explore unfamiliar APIs, figure out where the real data lives, and handle custom schemas, collapsing weeks of integration work into hours. If we get this right, it’s a genuine unlock to scale: no onboarding calls, no custom integration engineering.

Stack

Backend: Python, FastAPI, AWS
Frontend: TypeScript, React, Next.js