AlwaysThere

• 

  Logo

Inspiration

This past weekend, Ishan's grandmother passed away from complications of Alzheimer's disease.

In her final months, his family did everything we could — but we were doing it in the dark, and from far away. The information about her days was scattered across phone calls, text messages, relatives, and caregivers who each saw only a piece. She was eating a little less. Sleeping a little differently. A little less herself. Each change looked small on its own; it was only later, looking back, that we saw the pattern.

We often wonder whether a caregiver trained specifically in Alzheimer's and dementia care might have recognized those changes sooner — and whether noticing earlier could have helped us seek professional guidance before things became a crisis. We'll never know. But the feeling of being thousands of miles away, piecing together fragments and hoping we weren't missing something, is one no family should carry alone.

I almost didn't come to this hackathon. My dad encouraged me to show up and share our story. This is it.

An estimated 57 million people live with dementia worldwide today, and cases are projected to roughly triple by 2050 (WHO / Alzheimer's Disease International); in India alone, an estimated 8.8 million people over 60 are affected. Behind each of those numbers is a family like mine.

No family should have to look back after losing someone they love and wonder whether they missed an important sign. That's why we're building AlwaysThere.

What it does

AlwaysThere helps families support a loved one living with Alzheimer's or dementia — especially families who live in another city or country and can't always be there in person. Care for the person they are, even when you can't be there.

• Finds specialized caregivers. Search and match with caregivers experienced in dementia routine support, wandering, sundowning, mobility, nonverbal communication, and comfort care — with a plain-language explanation of why each caregiver is a good fit.
• Builds a Memory Care Passport. A person-centered profile that tells caregivers who someone is, not just their condition: routine, preferred name and language, familiar people and photos, favorite music, calming strategies, triggers, mobility and safety needs. It travels automatically to every caregiver — so a last-minute replacement already knows how to help.
• Records a baseline of "normal." Families note what's typical — appetite, hydration, sleep, mobility, mood, recognition — so the app can compare new observations against it.
• Tracks daily care. During each visit, caregivers log meals, hydration, medication reminders, mobility, mood, communication, sleep, and safety with quick checklists — no long forms.
• Turns it into a live timeline + daily summary. Families far away see verified updates in real time and a plain-language recap at the end of each day.
• Raises gentle, structured alerts. Green (informational), amber (worth a look), red (immediate safety) — always showing what was reported, when, by whom, and how it differs from the baseline.
• Speaks the family's language. Caregivers can leave a voice note in Marathi or Hindi; the family sees a transcript and English translation alongside the original.

What it explicitly does not do: diagnose, determine disease stage, predict outcomes, recommend medication changes, or replace a doctor or emergency service. It helps families notice meaningful changes earlier and communicate more clearly — nothing more.

How we built it

• App: React Native + Expo (TypeScript), Android-first because our users are in India — runnable on any phone by scanning a QR code in Expo Go.
• Backend: Supabase — Postgres, Auth, Storage, and Realtime (WebSockets) so a caregiver's update appears on the family's screen the instant it's logged.
• Privacy by design: Postgres Row-Level Security enforces access at the database — a loved one's home address is readable only by a caregiver with an active assignment, not just hidden in the UI.
• Concern detection is a deterministic rule engine, not an AI — it compares recent observations to the family-set baseline to raise green/amber/red. No model ever makes a medical judgment.
• Voice pipeline: Sarvam AI / Whisper for Marathi/Hindi speech-to-text, with translation to English; Google Gemini generates the plain-language daily summary and phrases alerts. The AI only phrases — the facts come from confirmed timeline entries.
• Everything runs on free tiers with no credit card — accessibility isn't just the mission, it's the architecture.
• Mock-first services mean the app runs and demos without any live keys.

Challenges we ran into

• Building something health-adjacent, responsibly. The hardest engineering wasn't code — it was deciding what not to build. No diagnosis, no disease staging, no predictions. Keeping the AI to phrasing and the alerts to transparent rules took real discipline.
• Real-time sync across two devices that feels instant and trustworthy.
• Indian-language voice — transcribing and translating Marathi/Hindi so a distant family truly understands what happened.
• Locking down sensitive data with Row-Level Security and keeping every API key server-side.
• Zero budget — engineering the whole thing on free tiers without cutting corners on privacy.
• Doing it while grieving — building this the same week we lost her was, honestly, the hardest part of all.

Accomplishments that we're proud of

• A real, runnable app — not a slide deck. You can scan a QR code and try it on your own phone.
• The living Memory Care Passport — turning a proven, paper-based idea (person-centered "life story" care) into a digital system that keeps care consistent across every caregiver.
• Verified updates in real time across two devices.
• Marathi/Hindi voice notes translated to English, so distance doesn't mean disconnection.
• A transparent, rule-based alert engine that helps families notice changes without ever pretending to diagnose.
• Built entirely on free, accessible tools.
• Most of all: turning a painful loss into something that might help another family.

What we learned

• Person-centered dementia care — capturing who someone is so caregivers can support them — is recognized best practice (like the Alzheimer's Society's "This is me" and life-story work). We learned it mostly lives on paper, and made it a living, connected system.
• Restraint is a feature. In a health-adjacent product, the most trustworthy choice is often to do less — and to say clearly what you don't do.
• Real, hands-on lessons in real-time architecture, database-level permissions, multimodal AI, and Indian-language speech.
• The elder-care space is real and crowded — our edge isn't "a marketplace," it's continuity of care and depth.
• How to channel grief into something constructive, together.

What's next for AlwaysThere

• Tap-to-Log NFC stickers — placed on the pillbox, kitchen, and door, so a caregiver logs a task with a single tap instead of typing, capturing exact timestamps in any language.
• Ambient sensor mesh — optional low-cost door, motion, and wearable sensors that feed the same baseline engine, watching over the hours between caregiver visits.
• A continuity network across diasporas — the model generalizes to any family caring for a parent back home, localized into any language pair by the AI translation layer, with distribution through partnerships with existing elder-care agencies and hospitals.
• A polished native build, real payments (UPI/Razorpay), and background tracking.
• Professional legal, privacy, and healthcare-data compliance review before any real family's data is ever stored.

AlwaysThere is being built so families can feel more informed, more connected, and less alone while caring for someone they love — even when they can't be there.

Built With

• alzhemiers
• android
• expo.io
• help
• ios
• react-native
• sql
• supabase