AI Engineer

We’re building one of the most ambitious AI products on the planet.

About Gravity

Hire AI agents built by top domain experts.

Great AI agents already exist. They’re stuck. The people who build them have no good way to earn from them, and everyone who wants one has to rebuild it from scratch. Chatbots answer or code. Builder tools hand you a blank canvas. So people pay agencies and freelancers, or try building it themselves and wonder why it isn’t giving results.

Gravity is building a platform where you hire AI agents that do great work. Experts build them. We handle everything else.

The role

This role is half AI product manager, half agent engineer. You will not hand specs to a dev team. You decide what problem to solve, design the agent behind it, build it with LangGraph, test it and ship it.

You work directly with the founder and own AI features from the first idea to production.

Fixed + Performance-based pay: you get the builder's cut of the revenue your agents earn, so your pay grows with what you ship.

What you will do

Product

• Break user problems down: understand each one deeply and split it into parts an agent can solve one at a time.
• Draw the lines: decide what an agent should do, what it should not, and where a human stays in the loop.
• Write it down first: clear requirements, success metrics and test scenarios.
• Learn from real usage: read the traces and failures, then decide what to fix next.

Agent building

• Build agent workflows with LangGraph and LangChain.
• Manage context and memory: what the agent remembers, forgets and retrieves.
• Write, test and improve prompts, with evaluations that measure their quality.
• Keep cost and latency in check with prompt caching, model choice and similar techniques.
• Connect agents to tools, APIs and data sources.
• Debug agent failures: loops, hallucinations, wrong tool calls, context overflow.

Ownership

• Ship fast, learn from users, iterate.
• Document what you build so others can understand and extend it.

What we are looking for

• You have shipped real products, not just prototypes, as a product manager, product engineer or similar.
• You have built with LangGraph or LangChain and can show it: a GitHub repo, a demo or a live product.
• You know how agents work: tool calling, planning, state, multi-step workflows, short and long-term memory, context limits.
• You measure quality and cost: tracing (LangSmith or similar), eval datasets, token cost and latency. You know an impressive demo is not a dependable product.
• You are comfortable in Python.
• You think and write clearly: a vague problem becomes parts, and the parts become a one-page plan.
• Most importantly, strong proof of work: agents you built, shipped, and fixed when they broke.

Good to have

• RAG, embeddings and vector databases in production
• Structured outputs, function-calling design and guardrails
• Several model providers, and knowing when to use which
• Multi-agent patterns and MCP
• Deploying and monitoring AI features: APIs, queues, observability, cost tracking, cloud basics

This role is not for you if

• You want to write specs and hand them to someone else to build.
• Your agents have only ever run in a notebook or a demo.
• You need a fully defined role before you start moving.

What you would get

• Real ownership from day one, working directly with the founder.
• A hand in core product decisions, not just tickets.
• The freedom to work from anywhere.
• The rare experience of working only with people who are excellent and kind.

We read every application and reply to the ones we connect with. If you are exceptional, Apply.