Staff Software Engineer

About the company

Our customer is a vertical AI company for the home services industry, serving HVAC, plumbing, electrical, and other trades businesses. Every one of their customers gets their own data environment, unified from the tools they already use and modeled into a semantic layer that agents can reason over.

Galaxy built and runs that data foundation alongside their team, and it’s live with real customers. Apply here and we’ll make a direct intro.

The Role

You’d be their first engineer, owning everything a customer touches. At the center is the AI harness, an agent runtime that takes a question, makes a plan, reaches into a customer’s data, and streams back an answer they can trust.

This is a product engineering role with AI at its core, not a research or model training role. The data infrastructure is covered, so you skip the plumbing and go straight to the product. You’ll work side by side with the founders, make your own architecture decisions, and help pick who joins next.

You will…

• Own the AI harness end to end, model agnostic across Claude, GPT, and Gemini

• Build evals that tell you when an answer is right and catch it when it gets worse

• Design a public API and an MCP server that bring the product into Claude, ChatGPT, and Gemini

• Build scheduled agents that run without anyone watching and deliver insights over email and text

• Ship the web app and chat experience a home services owner actually wants to use

• Keep tenant isolation, roles, and scopes airtight across every customer

You have…

• Shipped an LLM product real users touch, with agent loops, tool calling, streaming, and cost and latency in check

• Strong TypeScript, front to back, and some React

• Designed APIs other developers rely on

• Built on a semantic layer rather than raw SQL against the warehouse

• Security and multi-tenant instincts

• Built background jobs that run on their own, ideally with Temporal

• Real product taste, and you think about the business owner, not just the engineer

• A habit of shipping fast with AI coding tools, with no spec and full ownership

Bonus points if you…

• Have built agents over structured data, semantic layers, or text to SQL

• Have built an MCP server or integrations into Claude, ChatGPT, or Gemini

• Know your way around GCP, deployment, and infrastructure

• Have taken a product from zero to one before

Details

• Location

  • San Diego, Los Angeles, San Francisco, or remote

• Compensation

  • Competitive salary plus founding equity

• Interview

  • Expect to walk us through an agent you actually shipped, what broke, and how you knew it was working