Founding Engineer

Responsibilities:

• Email-native agents: Read submissions from shared inboxes, extract data from complex PDFs, identify missing information, email brokers for documents, etc.
• Context engine from historical data: Mine years of email threads to teach agents how each wholesaler works—which carriers they use, what documents they typically need, how they structure submissions, etc.
• Browser automation at scale: Automate data entry into niche carrier portals using browser automation while handling login/authentication.

This is what our stack looks like:

• Frontend: React
• Backend: TypeScript
• Database: Distributed SQLite
• Object Storage: Tigris Data
• Infrastructure: Fly.io
• LLM: OpenAI GPT models
• Coding Tools: Cursor, Claude Code and GitHub Copilot

Requirements:

• 2-5+ years full-stack experience building web applications, with at least 2 years focused on AI/ML-powered systems
• Have built applications that parse emails, extract attachments, handle MIME types, and interact with IMAP/SMTP or email APIs
• Have built production React apps with complex forms, file uploads (PDFs/docs), and real-time async updates
• Real-world experience with Elixir/Phoenix/Oban/Erlang/OTP

How We Think About Hiring For This Role:

Builds for the real world, not the ideal one

• You assume data is messy, systems break, and edge cases are the norm
• Example: PDFs come in rotated, handwritten, or as photos. You build extraction that handles this from day one, not after it breaks in production.

High agency prioritization and problem-solving. 

• You ship what matters most, not what's easiest or most interesting
• Example: You notice brokers attach wrong document types 30% of the time AND our PDF parser fails on rotated images 5% of the time. The parser is more interesting technically, but you fix the document validation first because it affects 6x more submissions.

Obsessed with details

• Good enough isn’t good enough when dealing with insurance
• Example: You catch that our agent extracted "$100,000" as "$10,000" once. You add checks, write tests, and make sure it never happens again.

Our Process:

1. First call with John (up to 30 min): You'll talk about the technical challenges you faced building a project you’re proud of. John will share what we're building and the challenges we’re facing. This is a two-way conversation - we want you to interview us too.
2. Paid practical exercise ($300, 2-3 hours): We'll give you real insurance documents and ask you to build something small that extracts and structures the data. Use any tools you want. We care more about how you think through the problem than perfect code.
3. On-site with Ola and John: Review your solution together, pair on extending it, discuss how you'd architect bigger systems. The three of us will grab coffee or lunch and we’ll make a decision within 24 hours.