Athos.

Senior iOS engineer on a biometric wearable — owning the full hardware‑to‑app pipeline. Sub‑second from sensor to coaching insight.

FIG. IV.A · Frontispiece EMG waveform
sensor → BLE → DSP → app
· live trace ·

Fig. IV.A A live EMG trace from the sensor garment, through BLE and on‑device DSP, into the iOS coaching screen.

§ 01The problem

Sensor data, useful in time.

Athos was a biometric athletic wearable: muscle activity sensors embedded in a garment, streamed wirelessly to an iOS coaching app. The product only has a reason to exist if the feedback loop is fast enough to act on — sub‑second from sensor to coaching insight, or it’s just an expensive shirt.

Most of the engineering problem was crossing layers: firmware, BLE protocol, on‑device signal processing, native client. A small drop in fidelity at any one layer is a noticeable degradation at the next one up.

Cross‑stack work is interesting because the bug is rarely where you start looking.

§ 02My role

Senior iOS, cross‑functional.

TitleSenior iOS Engineer · Real‑Time Systems

Tenure2020 — 2021

SurfaceFirmware → Protocol → DSP → iOS

ModernisedObj‑C → Swift / SwiftUI

The job nominally said “iOS engineer.” In practice I worked across the entire vertical because the product needed it.

Firmware. Embedded C on the wearable, including the BLE protocol layer.

Protocol. The BLE communication contract between sensor and phone, optimised for power and latency.

Signal processing. On‑device DSP turning raw sensor data into the metrics the app actually displays. Tuning for noise, dropouts, the lossy realities of BLE.

iOS. Native client redesigned around real‑time visualisation, long‑running sensor sessions, and the architectural patterns SwiftUI made newly viable. Modernised the platform from legacy Objective‑C to Swift / SwiftUI in parallel.

§ 03The stack

What it was built on.

FirmwareEmbedded C on the wearable sensor, BLE radio layer. C · BLE GATT · low‑power timing

ProtocolCustom BLE profile, framed transport, retry semantics. BLE · GATT · framing

DSPOn‑device signal processing of raw EMG. Swift · Accelerate · DSP filters

iOS / SwiftLive‑streaming client, sustained sensor sessions, real‑time visualisation. Swift · SwiftUI · Combine · Core Bluetooth

MigrationLegacy Obj‑C → Swift / SwiftUI rewrite in parallel. Objective‑C → Swift · SwiftUI

FIG. IV.B Sensor → DSP → screen
· vertical path ·

FIG. IV.C SwiftUI live
visualisation surface
· post‑migration ·

Fig. IV.B–C The sensor‑to‑screen path, end to end; and the SwiftUI visualisation surface that replaced the legacy Obj‑C client.

§ 04Outcomes

What shipped.

01Sub‑second latency. End‑to‑end from sensor to coaching screen.
02Cross‑stack ownership. Firmware, protocol, DSP and iOS in the hands of one engineer — fewer handoffs, faster diagnosis.
03Modernisation. iOS codebase migrated from Objective‑C to Swift / SwiftUI without breaking the shipped product.
04Session stability. Long sensor sessions handled without dropouts at the protocol layer.

Endnote

Working across the stack is the cheapest way to learn what a system actually does, as opposed to what it was supposed to.

— a habit picked up at Athos.