Harmony AI

Harmony AI - Helping the elderly reclaiming their independence

The Problem

Millions of older adults live at home, often alone or with an elderly spouse, quietly managing days filled with small but important needs: taking meds, making appointments, staying connected.

But they don’t always ask for help, and their families are busy, distant, and worried.

Existing tools (e.g., calendars, pill reminders, smart speakers) are fragmented and confusing. There’s no central solution that supports independence while keeping families informed.

The Market Opportunity

60M+ U.S. seniors, 85% living outside institutional care
63M Americans are family caregivers, mostly adult children balancing work and distance
The eldertech market is projected to exceed $120B in the next decade
Families are actively seeking tools to give them peace of mind that their elderly relatives are safe and healthy

Our Solution - Harmony

Harmony is a voice-first AI assistant for older adults and their families. For the elder:

Offers a daily summary: meds, events, weather, birthdays
Reminds, checks in, and notices when something’s off
Uses natural language: no screens, no effort For the family:
Sends automatic updates:
“Mom took her meds.”
“Dad missed a check-in.”
“Here’s a doctor summary.” It’s not just another app. It’s a quiet, reliable presence, supporting independence and peace of mind.

Why Now?

We’re at a generational inflection point:

A rapidly aging population
Caregiver burnout at historic highs
Tools that feel fragmented, cold, or inaccessible
And AI finally capable of listening, adapting, and understanding Harmony is redefining aging as an era of empowered independence, in harmony between elders and their family

Challenges we faced

Technical challenges

The biggest hurdle was connecting real-time speech-to-text and text-to-speech systems, ensuring low latency and natural flow for users who may rely entirely on voice.

Moreover, we had some additional features we wanted to include - such as sending texts to family members - that we were unable to complete in the time period. For example, sending texts to family members requires the Twilio API, which has a 7-day approval period. While we got Twilio to work, the actual SMS verification process would take longer than the Hackathon window.

Generally, we found that without technical experience, getting the code to work would have been really hard. We often had to go into the code to change it to achieve our desired functionality, and we found that understanding the fundamentals of OOP was really helpful to do so. Some things that are really simple engineering-wise - like adding a logo to the top-right of the app - proved to be more difficult than we thought simply through Lovable's chat interface.

Finally, versioning is not sophisticated, making it difficult to parallel process our work.

Prompt engineering challenges

We found the prompt engineering to be a bigger challenge than we initially thought. You need to be extremely specific with what you want, or Loveable's AI might misinterpret the prompt and make incorrect updates, wasting resources such as credits and time. One team member noted, "I used to laugh at the idea of prompt engineering as a job, but if you can be very good at this, it's definitely a useful skill."

What we learned

How to vibe code - We had no experience vibe coding before. While we don't think this will replace engineers, we gained an understanding for how AI will dramatically transform the way we build products. It's scary, but it's also cool.
The future of funding - Vibe coding also dramatically lowers the barrier for non-technical people to prototype. It will raise the bar for what's fundable. Less emphasis on the front-end, more emphasis on the back-end.
Kevin is insane