A consumer neurotech startup · Wearables & wellness

Making a brain-sensing headband useful: the app, backend and shop that sit behind it.

9:41SESSIONCALM / FOCUSDeep calm: 10 minBLE link: connectedSuggested: Body scanEEG SIGNALFilter · FFTON-DEVICELow latency
The brief, as we understood it
“
The headband reads brain activity while the user sits calm. The app has to turn that signal into meditation practices worth trying - in real time, and without lag.
- Our reading of the brief handed to Milleniance at the start of the engagement
01 - The challenge

A signal that has to become advice fast enough to feel live.

The headband measures the electrical activity of the brain while the user sits calm, and sends it to a phone over Bluetooth Low Energy. The app then has to recommend meditation practices that move the user toward a desired mental state. Because this is a real-time feedback product, delay is not a detail - sending every reading to the cloud and back could be a deal breaker.

The signal-processing logic also existed as Python scripts, which had to become native Swift and Java code. Around that core sat the usual weight of a consumer product: user accounts, content, device management, an admin portal - and a pre-order shop that needed to launch alongside it.

  • Cloud round-trip latency could undermine real-time feedback
  • Python signal-processing scripts had to be converted to native Swift and Java
  • App, admin tools and shop all needed to be ready for pre-orders
HEADBANDEEG sensingBLEPHONEsignal inON-DEVICEfast, works offlineCLOUD?latency risk
02 - The approach

Divide the system into features before writing any code.

The first task was a requirements document drawn from the hardware's components and interfaces. From there the design was divided into subsystems, treated as features, with the aim of keeping interdependencies between them as small as possible. That split makes it clear early which skills the build needs, what the software architecture looks like and where the bottlenecks are likely to be.

Two questions had to be settled before the rest: where signal processing should run - on the device or in the cloud - and which cloud provider to use. The team profile followed from the answers: native Swift and Java, on-device machine-learning kits such as Core ML and TensorFlow Lite, and a working grasp of filtering, FFT and spectrography.

  • Requirements drawn from the hardware components and interfaces
  • Features defined as independent subsystems to limit dependencies
  • Latency-critical processing location decided before build
FEATURES AS INDEPENDENT SUBSYSTEMSBLE LINKpairing, streamSIGNALfilter · FFTPERSONAL-ISATIONCONTENTmusic · playerACCOUNTSadmin portalSTOREpre-orders
03 - What we built

One companion platform. Four capabilities around a single signal.

Headband connectivity and signal pipeline
A Bluetooth Low Energy connection to the headband, with filtering, FFT and spectrography-based processing ported from Python prototypes into native Swift and Java, and incoming data passed to the server for further processing.
AI-personalised meditation
A personal database for every user and AI-driven personalisation, recommending practices organised by goal - relax, sleep, focus, compassion, energise and stress. Calm and focus trend charts, streaks and skill progress for users.
Music, content and player
Spotify, Apple Music and YouTube API integration, access to the user's own stored music, and a guided-meditation player alongside recently played and popular content.
Built on

A stack chosen for real-time response on the device, with the cloud doing what it does best.

Swift (iOS) Java (Android) Core ML and TensorFlow Lite Bluetooth Low Energy Cloud backend WooCommerce on WordPress
05 - What the scope covered

A full companion platform, scoped as independent subsystems.

2
Native mobile platforms, iOS and Android
3
Music and video services integrated
6
Goal-based meditation categories
7
Core platform capabilities, from database design to reporting

"If your product's value depends on a signal being interpreted in real time, decide where that processing lives before you design a single screen."

Let's see if we're the right fit

Thirty minutes with a senior product and IoT architect, before either of us commits to anything.

No sales deck, no discovery call with someone who hands you off afterwards. We look at your device, your data path and where the processing needs to happen. We respond within one business day.

Talk to a Senior IoT Architect →