50 long-form articles on embedded programming, hardware projects, edge AI, system design, DevOps, and the tools we keep coming back to. Written by working engineers. No hype, no listicles, no scraped tutorials.
Curated routes through the section, depending on what you are trying to learn.
From the constraints of an 8-bit microcontroller to the hidden costs of a cache on a Cortex-M7. The software that runs on the computers nobody thinks of as computers.
Start here. Conceptual foundations that the rest of embedded engineering builds on.
Most of the computers in your life are not the ones you sit in front of. They are inside your microwave, your car, your coffee machine, your insulin pump. Embedded programming is how those computers get told what to do.
02 EmbeddedThe debate is older than most firmware engineers. C is the historical default; C++ has quietly taken over huge parts of the industry. Here is when each one is actually the right choice.
03 EmbeddedAn RTOS sounds professional. Bare metal sounds primitive. Neither intuition is reliable. The right choice depends on concurrency, RAM, timing, and team size — and most firmware projects get this wrong in the same direction.
What happens between reset and main(), and where every byte of RAM goes.
Firmware that crashes in the field six months after shipping is almost always a memory bug. Here is how to reason about stack, heap, and static allocation on a microcontroller, and the patterns that prevent the disasters.
05 Boot ProcessWhat actually happens between pressing reset and main() running? Most embedded engineers go through their whole career without learning. Once you do, debugging boot issues, writing bootloaders, and customising memory layouts stop being magic.
The tools and techniques that turn long bugs into short ones.
Most beginners get stuck at printf debugging because the next step looks intimidating. SWD costs $5, sets up in ten minutes, and changes how you write firmware forever. Here is the honest tour of debug tools and when each is worth the setup time.
07 PowerThe difference between a battery-powered project that runs for a week and one that runs for a year is almost entirely about power management. Sleep modes, peripheral gating, wake sources, and the real numbers from the bench.
08 DMAMost CPU time in firmware projects is spent shuffling bytes between peripherals and memory. DMA does it without involving the CPU at all, which is why every serious embedded application uses it. Here is what it is, when to use it, and the gotchas.
Field-guide articles on protocols, RTOS primitives, and PCB design that you will refer back to.
Every embedded engineer needs fluency in I2C, SPI, UART, and CAN. They look interchangeable in tutorials and could not be more different in practice. Here is the side-by-side comparison with real use cases.
10 RTOSReal-time operating systems sound formidable. The vocabulary is the hard part; the concepts are mostly things any programmer already understands. Here is the practical introduction with FreeRTOS code.
11 ReferenceHTTP is the easy default. MQTT is the right answer more often than people realise, especially for battery-powered devices that report continuously. Here is the honest comparison with real packet sizes and battery numbers.
12 ReferenceDesigning a printed circuit board sounds intimidating until you have done it once. The workflow is short, the tools are free, and a manufactured five-board run from JLCPCB costs five dollars. Here is the field guide.
Arduino, ESP32, sensors, and the practical engineering of building things that measure, respond, and connect. Wiring diagrams, firmware patterns, and the gotchas that only show up at 2 a.m.
Picking the right hardware before writing any code.
The Arduino brand has grown from one board to more than twenty, and most online guides still recommend the wrong one. Here is the honest field guide: specs, prices, and what each board is genuinely useful for in 2026.
02 ESP32Espressif's chips quietly took over IoT. Between the original ESP8266 and the seven-and-counting ESP32 variants, picking the right one is genuinely hard. Here is the full field guide with a decision flowchart.
03 SensorsThe sensor you pick and how you wire it determines whether your project works reliably or crashes mysteriously after six hours. Here is the honest guide to the common sensors, their interfaces, and the gotchas nobody warns you about.
Techniques the first ten projects teach you. Save yourself the lessons.
End-to-end projects that combine sensors with WiFi, LoRa, ESP-NOW, or cellular.
A complete, working project: read a BME280 every 10 minutes, send the reading to a cloud backend over HTTPS, sleep in between, run for months on a battery. Architecture, firmware, and the pieces every tutorial glosses over.
06 ProjectWiFi stops at 30 meters and costs power. LoRa goes 2-10 km and runs for years on a pair of AAs. For a farm, a warehouse, or a large house, LoRa is the right radio. Here is how to build a network of three nodes and a gateway.
07 ESP-NOWESP-NOW lets two ESP32 boards talk directly to each other — no router, no IP, no internet. Five-millisecond latency, 250-byte packets, hundreds of meters range. Here is how to use it for projects WiFi cannot do.
08 ProjectA WiFi tracker stops working the moment the vehicle leaves home. A cellular tracker keeps reporting from anywhere with mobile coverage. Here is how to build one with an ESP32 and a SIM7600 module, including the cellular gotchas.
09 ProjectAn outdoor weather station with no power cable, no battery to swap, and no maintenance. ESP32, BME280, anemometer, rain gauge, a small solar panel, and the right power management. Install once, runs for years.
Projects that look at, listen to, or take care of your environment.
Anyone who has killed a houseplant by neglect can appreciate a system that just keeps it alive. Capacitive moisture sensor, small pump, an ESP32, and about twenty dollars. Here is the whole build with failure modes that matter.
11 ProjectIndoor CO2 rises faster than people realise. After two hours in a closed room it hits 1200 ppm; cognitive performance drops measurably above 1000. This monitor reads it in real time and tells you when to open a window.
12 ProjectA $12 ESP32-CAM wakes on motion, captures a photo, and sends it to your phone through Telegram. No cloud subscription, no monthly fee. Runs for weeks on a 18650 battery. Here is the full build including the gotchas that break it.
13 ProjectA $5 sensor and an ESP32 give you a working pulse oximeter. The hardware is easy; the signal processing is what separates a working build from a useless one. Here is how the MAX30102 actually works and code that produces sensible numbers.
14 ProjectThe line-following robot is the embedded equivalent of a programming bootcamp's calculator project — it teaches more about real-world control systems than any textbook. Here is the build, the code, and a simple PID that actually works.
Machine learning that runs on a microcontroller, not in the cloud. Wake words, person detection, anomaly classification — all in 16 KB of RAM and microamps of power.
From your first deployed model to vision and audio at the edge.
Running a neural network on a microcontroller sounded absurd five years ago. Now it is routine: the ESP32-S3 has 16 KB of TensorFlow Lite Micro runtime and inference under 100 ms. Here is how to actually deploy a model.
02 TinyMLA doorbell that recognises people instead of false-triggering on shadows. A trail camera that ignores rabbits. Person detection on an ESP32-S3 is genuinely viable in 2026 — here is the practical build.
03 TinyMLWake-word detection on an ESP32-S3 makes voice activation possible without sending audio to a server. No cloud, no monthly fees, no privacy compromise. Here is how to actually build it.
Architectural decisions that scale, break, and evolve. Covering consistency, caching, partitioning, and the tradeoffs the buzzwords paper over.
The vocabulary and theorems every backend engineer needs fluency in.
System design is the most over-taught and under-explained topic in software engineering. Here is what it actually covers, why it matters, and how to start thinking about it without memorising interview cheat sheets.
02 System DesignThe CAP theorem is one of the most misquoted results in distributed systems. Every time someone says "we chose CP over AP" they mean something different from the last person who said it. Here is what the theorem actually says.
The choices that shape how a system fails, scales, and evolves.
For ten years microservices were the default answer. The industry is now quietly walking that back. Here is how to decide which architecture fits your actual problem, without cargo-culting either direction.
04 System DesignCaching is the single highest-leverage performance technique in most backend systems. It is also the source of an alarming percentage of production incidents. Here are the patterns, the pitfalls, and the decisions that actually matter.
Designing the interfaces clients consume and the auth they pass through.
REST is the default; GraphQL is the trendy alternative; gRPC is the high-performance option. Each one is the right answer for a different shape of problem. Here is the honest comparison with code samples.
06 AuthAuth is one of those topics where everyone knows two patterns but rarely understands the trade-offs. Here is the field guide to sessions, JWTs, OAuth, OpenID Connect, API keys, and mTLS — what each is, when to use it, and what breaks.
Messaging and observability — the plumbing that production demands.
Three message queues dominate modern backends and they solve different problems. Kafka for high-throughput logs, RabbitMQ for flexible routing, SQS for managed simplicity. Here is the comparison and when each is the right pick.
08 ObservabilityLogs alone do not tell you why a request was slow. Metrics alone do not tell you which user was affected. Traces alone do not tell you what happened. Production observability is the discipline of having all three working together — and it is what separates teams that ship reliably from those that do not.
Containers, clusters, pipelines, and the boring infrastructure that makes the interesting code keep running. Honest takes on what to learn, what to skip, and when each tool is right.
Where most modern infrastructure starts.
Docker feels weird at first because it is genuinely weird. Once the mental model clicks, it becomes one of the most useful tools in software. Here is the introduction without the cargo cult.
02 KubernetesKubernetes is the most over-adopted technology in modern infrastructure. It is also the right answer for a specific class of problems. Here is the honest tour: when you actually need it, when you do not, and what it actually is underneath the YAML.
How code becomes a running system.
Continuous Integration and Continuous Deployment have become the default but most pipelines we see are slow, flaky, or unsafe. Here is the architecture for pipelines that work: stages, secrets, deploy strategies, and the antipatterns that destroy team velocity.
04 TerraformClicking around AWS to set up infrastructure works once and never again. Terraform turns infrastructure into code: versioned, reviewable, reproducible. Here is the field guide to getting started without the usual pitfalls.
The kit on the bench. Honest reviews of oscilloscopes, logic analyzers, PCB software, and the rest of the gear that separates a guess from a measurement.
Instruments that turn guessing into measuring.
Buying your first oscilloscope is overwhelming because every spec sounds important and every model claims to have the right amount of each. Here is the honest field guide that cuts through the marketing.
02 ToolsA logic analyzer is the single best return-on-investment in an embedded engineer's toolkit. The cheap clones at $10 are surprisingly capable; the genuine Saleae at $400 is a different league. Here is what to buy and why.
The tools that surround the bench: software, irons, microscopes.
KiCad has gone from interesting open-source project to the consensus choice for most engineers. Altium remains the industry incumbent. Eagle is dying. Here is the honest 2026 comparison and migration guide.
04 ToolsThe right tools turn a frustrating debugging session into a thirty-second job. Here is the equipment that earns its place on every embedded engineer's bench, organised by budget tier from $100 to $2000.