When Software Meets the Physical World
Exploring intelligent connected devices and IoT systems — mesh networks, telemetry ingestion, and the edge compute that turns raw sensor noise into decisions.
An IoT system is software that has to care about physics. Batteries drain, radios interfere, walls attenuate signals, and a sensor in a ceiling might not be reachable for two years. The engineering is less about the code and more about what the code can and cannot assume.
The mesh
A single gateway cannot cover a building. A LoRa mesh lets low-power nodes relay each other's packets, so a 900-node deployment needs only a handful of gateways. Each node sleeps for minutes at a time and wakes to transmit a few bytes.
{
"node_id": "n-0418",
"mesh": {
"protocol": "lorawan",
"band": "eu_868",
"spreading_factor": 7,
"battery_mv": 3120,
"hops_to_gateway": 2
},
"reading": {
"temp_c": 22.4,
"rh_pct": 48.1,
"co2_ppm": 612,
"ts": "2026-07-19T08:14:02Z"
}
}
Telemetry ingestion
Thousands of nodes produce a firehose of small messages. The ingestion path buffers, deduplicates, and writes to a timeseries store where downsampled views keep queries fast even after years of data accumulate.
-- average CO2 per floor, per hour, from raw 30s samples
SELECT
floor_id,
date_trunc('hour', ts) AS hour,
avg(co2_ppm) AS avg_co2,
max(co2_ppm) AS peak_co2,
count(*) AS samples
FROM sensor_readings
WHERE ts > now() - interval '24 hours'
GROUP BY floor_id, hour
ORDER BY hour DESC, floor_id;
The sensor is cheap. The decision it enables is not. Spend the effort on the path from reading to action.— IoT Mesh
The payoff is a building that knows itself — a live floorplan drawn from its own sensors, an HVAC loop that reacts to actual occupancy, a maintenance alert before a pump fails. Software, finally, touching the world it always described.