Junior AI Automation Engineer (Claude / Supabase)

About the Role

We close out construction projects: permits, inspections, correction notices, and the document package that gets a Certificate of Occupancy issued. Right now, a lot of that work is still done portal by portal, email by email, spreadsheet by spreadsheet.

We’ve built a single portal where customers can see everything in one place in real time, we run projects through it currently. You’ll be making direct changes to that portal and seeing your work show up in front of paying customers, fast.

We’re hiring a junior Claude Code developer to help build and automate this. This is a ground-floor role with profit sharing: you’re early enough that what you build will shape the product, the stack, and how we operate long term.

You don’t need construction experience. You do need to be comfortable reading messy real documents, turning them into structured data, and wiring that into simple tools and scripts that teammates and customers actually use.

You’ll spend part of your time helping run closeouts and part of your time improving the portal and automations that slowly replace that manual work. The target: a year from now this role handles 3x the project volume with the same headcount because of what you’ve built.

What You’ll Do

Own pieces of the closeout process (with support)

• Track permits and inspection status across multiple jurisdictions and portals
• Help schedule and confirm inspections; log results, correction notices, and re-inspections
• Assemble closeout packages: final sign‑offs, reports, waivers, warranties, O&M manuals
• Chase down missing items from subs, engineers, and inspectors before they become delays
• Flag blockers for project managers with the relevant info already pulled together

Build and improve our customer portal (Claude Code + Supabase + GitHub)

• Use Claude Code to write and refine scripts that log into portals, pull status, and normalize it into Supabase
• Call APIs where they exist; use light web scraping where they don’t
• Use LLMs (Claude) to extract structured data from inspection reports, correction notices, and permit PDFs
• Power the customer-facing portal with clean data so clients can see real-time status, issues, and next steps
• Build automated status tracking, alerts, and daily digests so nothing sits unnoticed
• Create basic dashboards (e.g., permit velocity, inspection pass/fail rates, aging docs)
• Work in GitHub for version control, pull requests, and code review so others can safely build on your work

Improve the system

• Instrument workflows so we can see where time is actually lost
• Decide where automation makes sense vs. where a better checklist is enough
• Document what you build so someone else can run it when you’re out
• Help evaluate new tools; make simple, honest calls on build vs. buy

What You Need

You’re early in your career (or just graduated) and want reps on real problems that touch real customers.

Technical

• Comfort with at least one programming language (Python preferred) and willingness to improve fast
• Basic SQL and familiarity with APIs, JSON, and HTTP requests
• Some experience using Claude or another LLM for coding / document processing (projects, coursework, or internships count)
• Interest in using Claude Code as your main development environment
• Experience using GitHub for version control and collaboration
• Familiarity with Postgres or Supabase (or a strong desire to learn; we use Supabase as our primary data layer)
• Git basics and the discipline to leave behind code someone else can run

Mindset

• You like taking vague, messy tasks and turning them into clear, repeatable workflows
• You’re careful with anything that touches regulators: you verify AI output, you don’t just trust it
• You’d rather ship a scrappy v1 this week and improve it than design something “perfect” for six months
• You’re motivated by seeing your work ship to real users and change how they work

Nice to Have

• Any exposure to construction, permitting, or inspections (even from family, side jobs, or internships)
• Experience with automation platforms (Zapier, Make, n8n)
• Basic web scraping, OCR, or computer-vision work on PDFs or scanned docs

How We’ll Evaluate You

We’ll do a practical exercise: you’ll get a stack of real (redacted) inspection reports and correction notices and we’ll ask how you’d get structured, reliable data out of them at volume using Claude Code, Supabase, and simple scripts. We care most about how you think through failure cases and verification, not whether you pick the “perfect” model or library.

Compensation & Benefits

• Salary: 45,000–55,000 (depending on experience)
• Profit sharing / upside tied to company performance
• Budget for AI tooling, API costs, and anything reasonable you need to build with*