← Back to site
← FRONT PAGE ASSISTIVE TECH PRICE: FREE — ALWAYS

The Saini Ledger

VOL. XXI · No. 8 SEATTLE, 2026 PAGE A2

ASSISTIVE TECH · ★ 4TH PLACE, NVIDIA TRACK

OFFLINE AI WATCHES OVER ALZHEIMER'S CARE

Local-first system watches over people living with Alzheimer's — before they know they need help

CORVALLIS, ORE. — People living with Alzheimer's often cannot recognize when they need help, so reactive assistance — which waits for a request — fails exactly when it matters most. Memory-care facilities are understaffed, and existing monitoring tools ship intimate behavioral data to the cloud. Linx, built in 24 hours at BeaverHacks and placed fourth in the NVIDIA track, takes the opposite approach: proactive, and fully offline.

The system runs on a low-cost hardware stack — a Raspberry Pi with a USB camera, an ultrasonic sensor, and a four-microphone array. Sensor readings flow into a rolling memory buffer, short enough to keep prompts lean and long enough to preserve context, then are reasoned over by NVIDIA's Nemotron 3 model running locally via Ollama.

HOW IT WORKS

A deterministic rule engine, dubbed Nemoclaw, runs in parallel with the model, evaluating known high-risk patterns — a stove left unattended, extended inactivity, a door left open — before inference even completes. Rules can escalate the model's risk assessment but never lower it, so safety-critical events are caught instantly and reliably.

Beyond alerts, Linx can notify emergency contacts, generate timestamped behavior reports for medical staff, and flag when nursing support is needed. The backend splits into six independent MCP microservices, which let the team swap a quantized model for constrained hardware or add a new sensor without touching the reasoning layer. Nothing leaves the device: zero data egress was a day-one constraint, not a feature.

LESSONS FROM THE FLOOR

For safety-critical detection, the team found a deterministic layer that runs before — and can only escalate — model inference more trustworthy than the model alone. The project was grounded in lived experience: having volunteered at an assisted-living center, the builders designed around how disorienting daily life with Alzheimer's can be. A concrete human need kept scope honest.

Source available → github.com/SNBest1/Linx ↗

CONTINUED: Model surfaces care disparities, Page A3 →