Skip to content
Blog

Notes on independence, privacy, and the machines that help.

How we build assistive technology that runs on-device — engineering deep dives, accessibility design, and news from the EchoSense team.

News

EchoSense on TV: Watan Raqami covers our graduation showcase

Watan Raqami featured EchoSense at the Arab Academy's graduation showcase — 'built at a quarter of the cost of international competitors, ready to become a startup.'

Read the article→
News

From graduation project to startup: the road ahead for EchoSense

EchoSense graduated from the Arab Academy with TV coverage and a working ecosystem. Here's our roadmap: pilots, local production, and affordable assistive tech at scale.

Read→
Accessibility

Designing for sensory calm: UI principles from the Companion app

Predictable grids, muted motion, user-picked themes, one-tap escapes: the sensory-calm design principles behind EchoSense's interface for autistic users.

Read→
Engineering

Building a wearable on Raspberry Pi 5 + Hailo: lessons from Echo 1

Bench-test every sensor, own every resource once, drop hardware that can't prove itself: engineering lessons from building Echo 1 on a Pi 5 with a Hailo AI HAT+.

Read→
Engineering

Running Gemma 4 on an iPhone with Apple MLX: what we learned

EchoSense runs a 4-bit multimodal Gemma 4 on-device via Apple MLX and mlx-vlm — a 7-layer prompt pipeline, output sanitizer, and tool parser. Lessons from production.

Read→
Engineering

Two triggers, zero single points of failure: designing SOS

EchoSense fires SOS from two independent paths — an AI tool-call and a deterministic life-threat text detector — so emergency dispatch never depends on one system.

Read→
Product

A caregiver dashboard built on consent, not surveillance

Live vitals, location, and SOS alerts for caregivers — granted by the user via QR + 6-digit linking, capped at 10 caregivers, revocable instantly. Here's the design.

Read→
Product

Breathing, stories, and check-ins: regulation tools that respect the user

Inside the Companion app's regulation toolkit — guided breathing, AI-narrated calming stories with persona picker, and a daily mood check-in built for sensory calm.

Read→
Privacy & Safety

How Echo remembers you without the cloud remembering you

EchoSense gives its AI companion long-term memory — interests, sensory notes, calming preferences — in encrypted on-device storage with semantic deduplication.

Read→
Engineering

Predicting distress before crisis with Apple Watch HR and HRV

How EchoSense turns Apple Watch heart rate and HRV into a personal, bounded stress score with four tiers — calm, elevated, stressed, crisis — without cloud processing.

Read→
Product

Meet Echo: an on-device AI companion for autistic users

The EchoSense Companion app runs a multimodal AI on your iPhone that learns your preferences, reads Apple Watch vitals, and guides calm — with total conversation privacy.

Read→
Accessibility

The real cost of independence: guide dogs, smart canes, and Echo 1

A trained guide dog costs tens of thousands of dollars; commercial smart aids run $2,000+. How EchoSense targets a quarter of competitors' cost with commodity hardware.

Read→
Engineering

Safety that never fails quietly: isolating hazard detection from AI

Echo 1's hazard detection is pure Python that never imports the AI model. Process isolation, audio priority, and two-layer restarts keep users warned no matter what crashes.

Read→
Product

Reading the world aloud: scene description and OCR on Echo 1

SmolVLM2 describes what's in front of you and Tesseract reads signs, labels, and documents aloud — on-device scene understanding for blind users, no cloud needed.

Read→
Engineering

An offline voice assistant with 13 tools and zero cloud

Echo 1's push-to-talk assistant transcribes with faster-whisper and answers with a tool-calling agent — time, weather cache, scene description, OCR — fully offline.

Read→
Accessibility

Why Echo 1 navigates with clock directions, not 'turn left'

Clock-direction cues like 'turn toward 2 o'clock' are the standard blind users already know. How Echo 1's route engine, earcons, and rerouting are built around them.

Read→
Engineering

Detecting drop-offs and steps with a single down-facing LiDAR

How Echo 1 uses a TF-Luna LiDAR, floor auto-calibration, and confirm-streak filtering to warn blind users about drop-offs, curbs, and steps before they reach them.

Read→
Engineering

How Echo 1 detects obstacles at 30 FPS on a wearable

Inside Echo 1's obstacle-detection pipeline: YOLOv8n on a Hailo-10H accelerator, clock-direction alerts, and severity arbitration — at 30 FPS, fully offline.

Read→
Privacy & Safety

Why on-device AI is the right foundation for assistive technology

Cloud AI fails exactly when assistive tech users need it most. Here's why EchoSense runs every model on-device — for privacy, latency, cost, and reliability.

Read→
Product

What is EchoSense? One wearable and one app for independence

EchoSense pairs Echo 1, an offline navigation wearable for blind users, with an on-device AI companion app for autistic users — private, affordable assistive technology.

Read→