MyHome AI

Inspiration

We were inspired by the everyday frustration of feeling helpless in your own home; staring at a confusing medical bill, a fridge full of random ingredients, a leaky pipe, or a party to plan with no idea where to start. We wanted to build one app that acts like a knowledgeable friend who can just look at your problem and tell you exactly what to do.

What it does

MyHomeAI is a five-feature AI-powered home assistant: it scans and audits bills and banking documents for overcharges, hidden fees, and fraud; then generates a negotiation script; it extracts event details from flyer photos and tells you when to leave based on your home address; it scans your fridge, identifies ingredients, suggests recipes, and builds a shopping list with local store prices; it diagnoses home damage from a photo and gives DIY repair steps, a parts list, and nearby professional contacts; and it plans parties by generating a full food plan with grocery store recommendations and cost estimates.

How we built it

We built the entire app in React with Tailwind CSS on the frontend, using Base44 as the backend; handling auth, the database, and serverless functions. Every intelligent feature follows the same loop: the user uploads a photo or file, a structured LLM call analyzes it with internet context enabled for real-world grounding, the result is saved as a database entity, and the UI renders the insights instantly.

Challenges we ran into

The hardest challenges were medical bill complexity (CPT codes, CDM rates, insurance underpayment), getting the banking fraud audit to run reliably in a single async backend function without timeouts, and making AI-generated store prices and travel times actually accurate by enabling live internet context rather than relying on model training data alone.

Accomplishments that we're proud of

We're proud that the entire app follows one clean mental model; point a camera, get an answer, and that genuinely complex tasks like fraud detection, bill negotiation scripts, and real-time grocery cost estimation all work end-to-end in a single user action with no manual steps required.

What we learned

We learned that strict JSON response schemas are non-negotiable for reliable AI pipelines, that feeding a home address to the model transforms generic answers into locally actionable ones, and that the hardest product decision is always what not to show — surfacing only the most useful information without overwhelming the user.

What's next for MyHome AI

Next, we want to add Google Calendar sync so events auto-populate, push notifications for "time to leave" reminders, a running monthly savings tracker across all audited bills, voice input for hands-free fridge scanning, and expansion into utility bill trend analysis so users can see month-over-month spending patterns and get proactive alerts when something spikes.

Built With

• css
• html
• javascript
• react
• swift
• tailwind