Hardware & Software — Deep Expertise Across the Full Stack

OGLAS isn’t a single-product company — it’s a full-stack engineering practice that designs, builds, and deploys across every layer of the embedded and server world. From 8-bit microcontrollers drawing microamps in a paddock to Rust cloud services running AI inference, we own the entire pipeline.

This page is a tour of the depth behind that claim.

Languages

We write production code in whatever the job demands:

• Rust — our cloud monorepo (seven products: Cloud, Local, Solar, Factory, Park, Debug, Marine), camera sidecar services, and increasingly new performance-critical work
• Python — AI/ML pipelines (PyTorch, SpeciesNet), camera capture and detection, data analysis, FastAPI services
• C — the OGLAS v2 binary codec (zero-malloc, no floats, integer-only wire format), Zephyr RTOS modules, display drivers
• C++ — the bulk of our PlatformIO firmware fleet across dozens of board targets, Arduino and ESP-IDF frameworks
• JavaScript — web dashboards (11ty/Eleventy), protocol twins, data visualisation (Chart.js)
• SQL — schema design, migrations, and query optimisation across SQLite and NATS JetStream

Frameworks & Operating Systems

We work across the major embedded and server frameworks — and choose the right one for each tier:

• Zephyr RTOS — our newest firmware tier, a west T2 workspace with devicetree overlays, Kconfig fragments, and native_sim testing via Twister. Sensor, Repeater, and Hub tiers all running Zephyr with over-the-air byte-identical compatibility to the legacy fleet
• FreeRTOS — the real-time kernel under ESP-IDF and Arduino ESP32, configured for our hub and display targets
• ESP-IDF v5.5 — native (non-Arduino) builds for our display fleet, using the ESP32-S3 LCD peripheral for RGB panels up to 1024×600
• Arduino + PlatformIO — our mature firmware fleet with dozens of board environments and a shared codebase
• Adafruit nRF52 BSP — low-power nRF52840 nodes with deep sleep down to 4 µA
• Docker — containerised cloud services, camera pipelines, and MQTT infrastructure
• nginx — reverse proxy, auth gateway, and TLS termination

Microcontrollers & Boards

We’ve shipped production firmware across four CPU architectures and over 25 distinct board targets:

ARM Cortex-M4 — nRF52840

Ultra-low-power sensor nodes and BLE + LoRa devices:

• RAK4631 WisBlock — primary sensor platform (nRF52840 + SX1262, modular WisBlock base)
• Seeed XIAO nRF52840 + Wio-SX1262 — compact companion radios and sensor nodes
• LilyGo T-Echo-Lite — GPS tracker with ePaper display
• SenseCAP T1000-E — multi-GNSS tracker (GPS + GLONASS + Galileo + BeiDou, LR1110 radio)
• SenseCAP Solar P1-Pro — solar-powered repeater with integrated GPS

Xtensa Dual-Core — ESP32 & ESP32-S3

Hubs, displays, WiFi + LoRa gateways, and trackers:

• LilyGo T3-S3 — hubs, repeaters, trackers, gateways, display panels
• LilyGo T-Deck — LoRa + WiFi display terminal
• Heltec WiFi LoRa 32 (V2, V3) — hubs, gateways, Victron solar monitoring
• Heltec Tracker V2 — GPS vehicle tracker with LoRa + WiFi + UC6580 GNSS
• Waveshare ESP32-S3-Touch-LCD-7B — 1024×600 RGB touch display
• Elecrow CrowPanel Advance 7" — 800×480 RGB display
• Waveshare RLCD 4.2" — reflective mono LCD (always-on, sunlight-readable)
• Waveshare Knob-Touch-LCD-1.8" — round QSPI display
• M5Tough — industrial panel (ILI9342 + AXP192 power management)
• M5Stack Dial — round colour display (GC9A01)
• AI-Thinker ESP32-CAM — WiFi camera nodes with OV2640

RISC-V — ESP32-C3 & ESP32-C6

BLE beacons and 802.15.4 Thread-capable nodes:

• Seeed XIAO ESP32-C3 — BLE + WiFi beacon and MQTT sensor
• Seeed XIAO ESP32-C6 — BLE + WiFi + 802.15.4 (Thread-ready)
• LilyGo T-Lora C6 — LoRa + 802.15.4 node

AVR 8-bit — ATmega328P

Legacy ultra-low-power sensor nodes:

• Talk2 Whisper Node — LoRa sensor with RFM95
• LowPowerLab Moteino — LoRa sensor with RFM95

Sensors

We integrate an enormous range of sensors — environmental, industrial, and specialised:

Temperature & Humidity

DS18B20 (1-Wire), TMP36 (analog), SHTC3 (I2C), BME280, BME680, SI7021, on-chip NTC thermistors

Pressure & Barometric

BMP180, BMP280, LPS22HB (RAK1902), BME280

Distance & Level

JSN-SR04T (waterproof ultrasonic, UART), RAK12007 (CS100), DYP-A02 family (A0221AU, A02YYUW — auto-stream UART)

Light

BH1750 (I2C lux), on-chip analog lux (T1000-E)

Motion, Tilt & Accelerometer

LIS3DH (RAK1904, 3-axis), QMA6100P (motion-interrupt wake for ultra-low-power), PIR (Panasonic EKMC, AM312, LD2420-OT)

Soil & Agriculture

RAK12035 capacitive soil moisture probe via RAK12023 I2C converter

Energy & Power

PZEM-004T v3 (Modbus-RTU CT clamp metering), Victron Energy MPPT (BLE encrypted telemetry decode)

Contact & Safety

Reed switches, DRV5032 hall sensors, float switches, leak probes, door/gate contacts

Pulse Counting

Hall flow meters (YF-B7), tipping-bucket rain gauges, reed anemometers (ISR pulse counters)

GPS / GNSS

RAK12501 (Quectel L76K), RAK12500 (u-blox ZOE-M8Q), UC6580, AG3335 (Airoha, multi-GNSS), L76K — all via TinyGPSPlus

Displays

We drive displays from tiny 0.96" OLEDs to 7" capacitive-touch RGB panels:

Touch controllers: GT911, CST816D, FT6336. GUI toolkits: LVGL (v8 and v9), Adafruit GFX, LovyanGFX, Arduino_GFX.

Networking & Radio

This is where OGLAS really stands apart. We don’t just use radios — we design protocols, build mesh networks, and decode manufacturer-proprietary BLE:

Wireless Protocols

• LoRa (sub-GHz, AU 915–917 MHz) — point-to-point with AES-CTR encrypted binary frames, our primary long-range transport
• LoRa-Basic — our fleet protocol: raw LoRa, encrypted, ~99% of deployed nodes
• MeshCore — encrypted mesh routing with store-and-forward repeaters
• LoRaWAN — feasibility assessed and documented
• BLE 5.0 — advertising beacons, observer mode, config + OTA DFU, manufacturer-specific decoding
• BTHome v2 — open BLE sensor protocol (WS90 weather station decode)
• Victron BLE — AES-CTR decryption of proprietary Victron MPPT solar charger telemetry
• WiFi (802.11 b/g/n) — hub uplinks, web UIs, cloud push, OTA updates
• 802.15.4 — Thread-capable nodes on ESP32-C6
• HaLow (802.11ah) — long-range WiFi for remote site backhaul

Radio ICs

SX1262, SX1276, SX127x, LR1110, RFM95 — Semtech and HopeRF LoRa transceivers across every band and form factor

Wired & Serial

UART (GPS, Modbus-RTU, companion links), I2C (sensors, OLED, touch), SPI (radio, SD, displays), 1-Wire (DS18B20), USB CDC (console, DFU)

Network & Application

• MQTT 3.1.1 — sensor ingest with QoS 1, TLS/mTLS capable, via NATS JetStream
• HTTP/HTTPS — REST APIs, web dashboards, cloud push, mbedTLS on ESP32
• RTSP — camera live streaming with H.264 fragmented MP4 via MediaSource
• ONVIF WS-Discovery — IP camera auto-discovery
• DNS-SD / mDNS — hub and device discovery (_oglas._tcp)
• SSE — Server-Sent Events for real-time AI chat and camera event feeds
• OTA — WiFi firmware updates (ESP32), BLE DFU (nRF52)

Low Power & Battery Expertise

We’re experts at making devices run for years on battery:

• nRF52840 WFE deep sleep — ~4 µA floor with GPIO wake, radio sleep, switched sensor rail
• Duty-cycled wake/publish/sleep — sensor nodes that wake, measure, transmit, and sleep in milliseconds
• Switched 3V3_S rail — power-down all peripherals during sleep
• QSPI flash deep-power-down — 0xB9 command for external storage
• GPS duty divider — configurable fix intervals for battery savings
• V_BCKP retention — GPS warm/hot starts across power cuts
• Beacon mode — fire-and-forget, no ACK camping for maximum battery life
• Motion-adaptive cadence — accelerometer interrupt wakes only on movement (T1000-E)
• Solar harvesting — SenseCAP Solar nodes with integrated panel

GPS & Tracking

Deep GPS expertise across multiple module families and use cases:

• Vehicle trip logging — persistent back-walk retransmission for complete trip replay
• Real-time position tracking — breadcrumb trails with LoRa position reports
• Solar panel tracking — GPS + SD logging for mobile asset monitoring
• Multi-GNSS — GPS + GLONASS + Galileo + BeiDou on T1000-E
• GPS fix optimisation — timeout management, catch mode for warm starts, configurable fix intervals
• TinyGPSPlus — NMEA parsing across all module types

Server & Cloud Infrastructure

Our server stack is as deep as our firmware:

OGLASCloud — Rust Monorepo

Seven products sharing a common architecture:

• Cloud — multi-tenant cloud store, push-ingestion API, dashboards
• Local — single-site software hub with serial/MQTT/hub-pull ingest
• Solar — multi-site power and solar monitoring
• Factory — plant overview with OEE, andon, uptime tracking
• Park — satellite maps, trail cameras, photo review for parks and reserves
• Debug — field debug and discovery tool for LoRa bandwidth monitoring
• Marine (WII5) — wave-buoy deployments and capture spectra

Shared Crates

oglas-core (data model), oglas-ui (dashboard canvas), oglas-wire (wire protocol), oglas-agent (AI/LLM integration), oglas-overlays (GeoJSON maps)

Data Infrastructure

• NATS JetStream — durable, replayable, immutable message log with MQTT gateway
• SQLite — local-first storage with migrations, full-text search, and geospatial queries
• 11ty (Eleventy) — static site generator for all web frontends
• Chart.js — data visualisation and dashboards
• Sh3dAuth — external magic-link authentication service

Data Ingest Paths

LoRa sensors → Hub → Cloud push · BLE sensors → Hub observer → Cloud · WiFi MQTT → NATS → Rust consumer → Cloud · Camera → Python service → Cloud ingest · Local hub → Cloud push

AI & Machine Learning

We integrate AI directly into the monitoring pipeline — not as a bolt-on, but as a first-class sensor:

• SpeciesNet (Google CameraTrapAI) — MegaDetector for animal/person/vehicle detection + EfficientNetV2-M species classification
• Australian geofencing — species filtering tuned for Australian wildlife
• Predator alerting — fox (Vulpes vulpes) and feral cat (Felis catus) detection with confidence scores
• Scene analysis — brightness measurement, monochrome detection (IR vs white light discrimination)
• Software motion detection — frame-difference algorithm for cameras without native motion API
• Motion burst — rapid multi-frame capture on motion events for higher classification confidence
• LLM integration — AI agent with SSE-streaming chat interface for natural-language data queries

Camera Support

Six camera driver families with unified processing:

• Dahua — HTTP Digest auth, motion events, dual-light control (IR/white/auto), SMD management
• HTTP snapshot — generic IP camera polling
• RTSP — ffmpeg-based streaming with live restreaming (H.264 fMP4 + MJPEG fallback)
• ESP32-CAM — custom firmware for WiFi camera nodes
• FTP folder-watch — legacy camera integration
• USB webcam — V4L2 capture
• ONVIF WS-Discovery + mDNS + TCP fingerprint for automatic network camera scanning
• Multi-camera concurrent processing with shared AI model and shared SQLite store

Security

• AES-CTR encryption on all LoRa frames
• ed25519 cryptographic signatures (MeshCore)
• MQTT TLS/mTLS for cloud-direct connections
• Token-based ingest authentication
• NVS credential storage — WiFi, MQTT, LoRa keys persisted securely
• BLE DFU — encrypted over-the-air firmware updates

Device Roles

Our firmware fleet covers every role in a monitoring system:

The OGLAS Protocol

At the heart of everything is the OGLAS v2 binary protocol — a transport-agnostic codec that knows nothing of LoRa or TCP:

• Forward-compatible field IDs with width-in-id scheme
• Integer-only wire format (no floats, no malloc)
• Single source of truth (oglas_fields.def) generates C and JavaScript bindings
• Verbs: sensor, beacon, ack, cmd, cmdresp, alert, node, have, pull, relay
• Byte-identical over-the-air compatibility between Arduino, Zephyr, and Rust implementations

This breadth isn’t theoretical — it’s deployed, in production, on real sites across Australia. If you’ve got a monitoring problem that spans hardware, radio, server, and AI, we’ve probably already built something adjacent to it.

Talk to us about your project →