From chipto cloud.

Layer 1

Embedded and BLE

Soulbond, a BLE haptic wearable I designed and built solo: the board, the firmware, the case, the app and the server behind it.

• Custom 4-layer PCB in EasyEDA Pro, nRF52840 with an nPM1300 power chip
• Zephyr RTOS firmware on a custom board definition
• BLE GATT services with over-the-air firmware updates
• Capacitive touch, haptics and I2C sensor drivers
• 3D-printed enclosure designed in Fusion 360
• Compose Multiplatform companion app and a Ktor backend
• Hands-on bring-up: soldering, breadboards and SWD debugging
• PCB design, firmware and CAD, self-taught in months

Layer 2

Native Android

Kotlin, Jetpack Compose, Coroutines and MVI, running in production apps.

• Led the migration to Compose from Epoxy and XML, with MVI
• Adaptive layouts for phone, foldable and tablet
• Custom Canvas animations in Compose
• Multi-module Clean Architecture, data, domain and presentation per feature
• Legacy Java modernized to MVVM, single activity and Jetpack
• Tested with unit, integration and UI tests
• Code reviews across projects and weekly Android sessions I teach

Layer 3

On-device ML

Real-time vehicle and plate detection on the phone itself, with no server round trip.

• TensorFlow Lite on CameraX frames
• Vehicle and licence plate detection
• Built for a driver-safety app

Layer 4

Android Automotive

Road warnings delivered to the driver on the car's own screen, as the sole Android developer for intelligent transport systems.

• Android Automotive OS head units
• Android Auto projection
• Radar-detected pedestrian warnings, delivered in-car
• Google Maps and Mapbox with real-time MQTT and SignalR
• City safety app: live location of every road user, SOS alerts and hazard warnings

Layer 5

Telecom and security

Telephony that goes deep into the Android stack, protected by layered app hardening.

• SIP calling in the native dialer
• Call and SMS interception with CallScreeningService
• Call recording and audio routing
• Secure XMPP messaging on a closed, encrypted network
• Device telemetry: signal, battery and GPS with a multi-level fallback
• OTA app updates with remote version checks

Seven layers of hardening

• RASP: root, debugger and emulator checks
• Frida detection in native C++
• Hardware-backed Keystore, AES-256
• SQLCipher database
• R8 and ProGuard obfuscation
• String encryption
• Certificate pinning

Layer 6

Multiplatform and backend (KMP/CMP)

One shared Kotlin core across Android and iOS, with type-safe API contracts to my own Ktor server.

• Android
• iOS
• Ktor server, REST and WebSockets
• Webcoming soon

Layer 7

Docker, self-hosting and IoT

I run my own infrastructure, from containers to the sensors that report to it.

• Multi-stage Docker images for Ktor services
• Compose stacks with MySQL and WebSockets
• Raspberry Pi, NAS and cloud VM working as one network
• Private DNS, my own certificate authority, reverse proxy and VPN mesh
• ESP32 sensor nodes reporting temperature and humidity over HTTP

Layer 8

AI orchestration

I don't just use AI. I orchestrate it: agents that build, review, debug and help me learn new domains fast, with me deciding what ships.

• Build: the iOS port of GPA Manager
• Review: pull requests
• Debug: firmware
• Learn: PCB design and firmware, self-taught
• Design: the wearable's enclosure
• Create: the showreel on this page

Layer 9

Quality and performance

Measure first, then optimize, then verify the numbers held.

Profiled with Android Studio Profiler, Perfetto and LeakCanary.