JG—26← All workContact ↗

08 / Project

Living Systems · IoT · Independent R&D

PermaTech

Environmental Intelligence Platform

Connecting physical environments with digital intelligence.

PermaTech is an experimental environmental monitoring platform at the intersection of physical computing, data engineering and interactive digital experiences.

I designed it end to end, from embedded firmware and wireless communication through data storage, real-time APIs and a custom visual interface. It explores how physical measurements can help us understand the living systems around us.

My role
Concept, UX/UI Design, System Architecture, Full-stack Development, IoT Integration and AI-assisted Engineering
Year
2026 — Ongoing
Project type
Experimental
Categories
Creative technology / IoT / Full-stack development / Data visualization
Areas
Full-stack development / IoT / Data visualization
Technologies
ESP8266 · C++ · MQTT · Node.js · TypeScript · PostgreSQL · Drizzle ORM · Express · WebSocket · Next.js · React · Tailwind CSS · TanStack Query · Zod · Docker · Jest · Playwright
Living Systems dashboard with physical DFRobot Lark selected, five current weather readings and recorded observation history, with simulated sources labelled separately
Living Systems brings current conditions, observation history and connected sources into one view.

01 / Context

Understanding living systems

Temperature, humidity, atmospheric pressure and wind change together. A reading captures a moment, but it takes a series of observations to see how conditions evolve.

PermaTech grew out of an interest in environmental monitoring, regenerative practices and connected devices. The challenge was to design a system that could collect physical observations, process them reliably and make them straightforward to explore.

02 / Hardware and firmware

Physical computing

The first physical instrument is a DFRobot Lark weather station connected to an ESP8266 microcontroller. Custom C++ firmware reads the station over I²C, normalizes the readings, synchronizes capture time using NTP and publishes structured telemetry over Wi-Fi. A unique capture identifier follows each observation through the system.

The station measures air temperature, relative humidity, atmospheric pressure, wind speed and wind direction.

Actual DFRobot Lark station wired to an ESP8266 on a breadboard beside the Living Systems dashboard
The DFRobot Lark and ESP8266, connected over I²C, alongside Living Systems.

03 / System architecture

From physical environments to digital systems

Physical and simulated devices publish environmental observations over MQTT. A Node.js service validates incoming messages and stores them in PostgreSQL. The versioned REST API serves latest observations and history, and WebSocket notifications follow successful database commits.

Acquisition, persistence and presentation are separate parts of the system. Other applications can consume the same environmental data independently of the Living Systems frontend.

Telemetry architecture

01 / Acquisition

Physical environment

DFRobot Lark → I²C → ESP8266
NTP capture time · UUID

Aedes broker

Physical QoS 0
Four labelled simulators · QoS 1

02 / Persistence

Validate & ingest

Node.js · TypeScript · Zod
Shared telemetry contracts

PostgreSQL

Message + five readings
Unique identity · commit

03 / Read & notify

REST API

Reads from PostgreSQL
Latest · History · Recovery

Only after commit

WebSocket

Node.js ingestion callback
Committed event · Best-effort

04 / Presentation

Living Systems

Next.js · React
Current readings · History

Independent clients can consume the same read-only API and event stream.

Acquisition, persistence and presentation remain separate. REST queries the stored observations. The ingestion service emits a WebSocket notification only after its PostgreSQL transaction commits; the database does not broadcast the event directly.

04 / Data engineering

Engineering the data lifecycle

Shared contracts validate incoming telemetry before an atomic database transaction stores the message and its readings. Unique message identifiers prevent duplicate persistence. Capture time records when the device measured the environment, and receive time records when the system accepted the message.

PostgreSQL serves bounded time-series queries and applies metric-specific aggregation. Missing measurements stay absent and zero remains a valid reading. Scalar metrics support minimum, maximum and average. Wind direction is circular, so each aggregated bucket reports its latest captured bearing.

Physical observation / lark-001

One observation. Four views.

Captured 19 November 2026, 10:26:36 UTC. The same identity and five values follow the observation through the backend.

Observation UUID

64267b73-6e4c-44be-95a0-75ebef87bbe0
Air temperature
21.24 °C
Relative humidity
73.14 %
Pressure
951.31 hPa
Wind speed
0 m/s
Wind direction
0 deg

The original non-retained MQTT publication from physical lark-001, delivered at QoS 0.

{
  "schemaVersion": 1,
  "messageId": "64267b73-6e4c-44be-95a0-75ebef87bbe0",
  "deviceId": "lark-001",
  "timestamp": "2026-11-19T10:26:36Z",
  "readings": [
    {
      "metric": "air_temperature",
      "value": 21.24,
      "unit": "°C"
    },
    {
      "metric": "relative_humidity",
      "value": 73.14,
      "unit": "%"
    },
    {
      "metric": "atmospheric_pressure",
      "value": 951.31,
      "unit": "hPa"
    },
    {
      "metric": "wind_speed",
      "value": 0,
      "unit": "m/s"
    },
    {
      "metric": "wind_direction",
      "value": 0,
      "unit": "deg"
    }
  ]
}
Download original JSON ↓

05 / Frontend engineering

A real-time digital experience

The Next.js and React frontend shares TypeScript and Zod contracts with the data service. A persistent WebSocket connection brings in newly committed observations, while REST supplies authoritative state, historical queries and recovery after a disconnect.

The client deduplicates events and reconciles readings by metric and capture time. Delayed observations cannot overwrite newer readings, and partial observations preserve the last known values of absent metrics. The current application supports physical and simulated devices, live measurements, historical exploration and telemetry freshness indicators.

The weather station in operation, with sensor readings flowing through the connected system.

06 / Visual language

Living Systems

The interface draws on solarpunk, scientific instruments, botanical illustration and organic generative forms. Deep petroleum blues, organic greens and sunlight tones frame the data, with parchment tones and editorial typography for reading.

The visual language is evolving toward a family of interactive instruments that help people notice patterns and relationships in environmental data.

Desktop device detail with physical Lark readings, observation history and source navigation
Desktop / A focused view of each source, its current readings and observation history.
Mobile overview with stacked metric readings, observation history and physical source labels
Mobile / The same visual language adapts to a smaller screen. Scroll to explore the layout.

07 / Verification

Engineering reliability

116 automated tests cover data contracts, telemetry ingestion, transactional persistence, API queries and frontend behavior. A physical observation can be traced across MQTT, PostgreSQL, WebSocket and REST through its unique identifier.

I led the architecture, product design, physical integration and verification, using AI to support implementation and testing. The client reconciles incoming observations and recovers its authoritative state through REST after a disconnect.

116
Automated tests passed
5
Environmental metrics
10 s
Configured sampling cycle
UUID
Identity across layers

Application baseline / 116 tests

41 contracts + 48 backend integration + 21 frontend data + 6 Chrome browser checks.

Physical evidence session / 110 tests

41 contracts + 48 backend integration + 21 frontend data, rerun on 19 November 2026.

08 / Scope and next steps

An evolving platform

PermaTech is ongoing independent research and development. Future work includes the Environmental Clock, final Device Detail visual refinement, radial visualizations, low-power and LoRa sensor networks, solar-powered instruments and applications in environmental and astronomical monitoring.

Next project
Morae →