Event
LA Hacks
Team
Snehanshu Raj, Panyi Wang
Memex.AI is a two-component assistive system built at LA Hacks 2026 for dementia and Alzheimer's patients who can no longer reliably use screens, apps or touchscreens. Built by Snehanshu Raj and Panyi Wang, Memex.AI tackles one of caregiving's most overlooked problems: the people who need technology the most are often the ones least able to use it.
What it solves
Dementia and Alzheimer's patients may not recognize who just entered the room, sometimes not even their own family members. They forget to take essential medication. Most assistive technology assumes a level of independence that simply is not there. Memex.AI is built for patients who cannot tap, scroll or navigate a screen, running entirely offline with no internet required.
What it does
A wearable device worn by the patient continuously scans for faces and announces who is in the room by name and relationship. It listens for a wake word, answers spoken questions, accepts voice commands, and speaks medication reminders at scheduled times through earphones. A caregiver iOS app enrolls people with 1 to 5 photos, reads prescription photos to extract structured medication data, and gives doctors an instant structured medication summary at every visit. Everything works offline with no screen interaction required from the patient.
How it works
The wearable device runs on a Rubik Pi 3 with InsightFace for face recognition, openWakeWord for wake word detection, Whisper tiny for speech-to-text, and Kokoro TTS for voice responses, all running locally offline. The iOS app uses ZETIC Melange to run Gemma 4 on-device via NPU for medication structuring from prescription OCR, and MediaPipe for face processing. The two devices connect over an encrypted Tailscale peer-to-peer network. No patient data is ever sent to the cloud.