CleanHaus
CleanHaus took its app and AI agent live, our engineers inside its team
CleanHaus runs a live platform for short-term rental turnovers. Our AI engineers grew its operations assistant and our mobile engineers built its new React Native app, both inside CleanHaus's own repositories, AWS accounts and logins.
Real Estate & HospitalityTurnover operations SaaSAI added to a live product
The challenge
CleanHaus had a live product, a first AI assistant and an aging mobile app, and needed engineers who could change both without slowing the product down.
What we did
AI and mobile engineers from Datics worked as part of CleanHaus's team, in its repositories and accounts, reviewed by its CTO and shipped through its own pipelines.
The outcome
CleanHaus reported its app and AI agent live and credited the Datics team, and the work continues inside its team.
- Model
- Datics engineers inside CleanHaus's product team
- Workstreams
- AI operations assistant and React Native app
- Runs in
- CleanHaus's own AWS accounts and logins
- Status
- App and AI agent live, work ongoing
The starting point
CleanHaus coordinates short-term rental bookings, cleanings, checklists and supplies between property owners, cleaner managers and cleaners, on a serverless AWS backend with web and mobile apps.
What had to change
Its AI assistant covered only part of the job cycle, and its Ionic and Vue mobile app needed replacing. CleanHaus wanted engineers who would work in its own codebase and process rather than hand over a separate system.
What could not break
- Code ships into CleanHaus's own repositories through its reviews
- AI acts only through CleanHaus's APIs and Cognito logins
- Builds ship through CleanHaus's own store accounts
- The product stays live while the assistant and app change
The calls we made
We worked as part of CleanHaus's team, not as a vendor beside it: its repositories, its accounts, its reviews.
- 01
Their accounts, not ours
Access was granted by CleanHaus's CTO to its own GitHub, AWS and Cognito, with separate development, experimental and production environments, so nothing ever had to be handed over.
Ruled outBuilding on our infrastructure and migrating at the end.
- 02
Show work as it lands
We met the CTO regularly, shared demo recordings as features landed and took designs to him before building, so the team saw each change before it reached users.
Ruled outLong stretches of work revealed only at a milestone.
- 03
Evals on every push
Tests run in tiers in CleanHaus's GitHub Actions, from unit tests on every push to LLM-as-judge evals, so a model upgrade became a routine change for its team.
Ruled outManual spot checks before releases.
How the system fits together
What we built
The AI operations assistant
Role-aware agents that act through CleanHaus's APIs under each user's login, reachable in app chat, by text, in Slack and by voice, with contractor outreach waiting for the owner's approval.
The new mobile app
One React Native codebase replacing the Ionic and Vue app, with phone sign-in, push actions, deep links and role-based calendars.
Tests and traces
Tiered tests and evals in CleanHaus's CI, and agent runs traced in LangSmith and CloudWatch so bugs are found from logs.
A closer look
What changed
CleanHaus reported its app and AI agent live and credited the Datics team. The assistant and the app both run in its own accounts and repositories, built by engineers working as part of its team.
“CleanHaus has been moving quickly to deploy our app and Broomy AI Agent with the key assistance of the Datics team. We couldn't have reached this point without your excellent bench of developers.”
Under the hood
- Datics role
- Forward deployed engineering
- Access
- CleanHaus's GitHub, AWS and Cognito, granted by its CTO, with separate development, experimental and production environments.
- Delivery
- GitHub Actions with tests on every push for the assistant, and Expo EAS builds through TestFlight for the app.
AI and models
Cloud and infrastructure
Apps and APIs
Also in the stack
How it shipped
Extend the assistant
We studied the running assistant, scoped its extension and started adding tools, channels and agents.
Build the app
Five planned sprints delivered the new React Native app's foundation and calendar beside the AI work.
Into production
Production environments, tiered evals and a multi-channel rebuild of the assistant.
Live and growing
CleanHaus took its app and AI agent live, and our engineers kept shipping inside its team.
In CleanHaus's process
Our engineers shipped through CleanHaus's pull requests and reviews, in its Slack channels, with its CTO setting priorities.
Demos as features landed
Recordings and live walkthroughs let CleanHaus's team see each feature before users did.
Knowledge stays with the client
Runbooks, tests and traces live in CleanHaus's own repositories and accounts.
What this unlocked
Next are voice calls for hosts and a plan for accounts with thousands of properties and users who hold several roles.
Related work
Have a product with the same kind of constraints?
Tell us what already exists and what you want it to do. We will tell you what it takes.