HOAi Engineering

We build AI agents that do the work of community management: they process invoices, answer owner calls, handle violations, and close the books. Current run rate: 9,000,000+ tasks a year.

$ a note to candidates

Engineers here direct fleets of coding agents. If that sounds like fun, we should talk.

How we build agents

Every agent gets its own computer

Each interaction runs on its own isolated machine, so the agent is free to run code, call tools, and write to systems of record.

Evals decide what ships

Community management had no benchmark, so we built one from real scenarios. An agent ships when it passes.

Every production trace becomes a test case

Humans review the agents' work in production. Each reviewed trace joins the eval set, and each caught mistake becomes a regression test.

  1. A board, resident, or manager request comes in.
  2. The agent runtime orchestrates it: planning, retries, idempotency.
  3. Work fans out to sandboxed tool execution, the state and memory store, and integration adapters.
  4. Observability covers every step: tracing, SLOs, incident response.

In production

tasks per year, current run rate 9,000,000+
invoices processed per month, one customer 15,000+
reduction in AP processing time, one customer 95%
reduction in budgeting time, one customer 90%
owner-call hold time, one customer 4 min → seconds
staff hours returned in 90 days, one customer 1,400+

Open roles (6)

Live from our careers board.

all roles at Vantaca + HOAi →

Interview process

  1. Intro

    We talk about what you've built and what we're building.

  2. Technical

    You solve coding and systems design problems that came out of our production work.

  3. Team session

    You work through a real problem with the engineers you'd join.

  4. Decision

    You end with a founder conversation. We come back with a clear yes or no.

Blog

all posts →

Interested?

Apply to a role above.