Last updated: August 24, 2026
By Robert Ardell, Co-Founder and Strategic Advisor, KORE1
An embedded software engineer job description has to name the silicon, the layer of the stack the person owns, where the hardware physically sits, and any safety standard the product is certified against. Base bands run $140,000 to $225,000 in 2026 depending on which of those four answers you give.
Half the embedded products shipping this year will still be under support in 2036. The Linux Foundation put a number on that in its Zephyr ten-year adoption report: 52% of organizations building on Zephyr support those products for five to ten years or longer. Ten years. The person who takes this seat is writing code that outlives the posting, the hiring manager who approved it, and almost certainly the board revision it was written against.
So the posting matters more here than it does for web work, and most embedded postings I read do the same thing. They open with “C/C++” and “RTOS experience” and a bullet about working cross-functionally with hardware teams. Then they stop. A firmware engineer reads that and learns nothing. Four genuinely different engineers can answer to every word of it, and three of them will fail your technical screen for reasons that have nothing to do with talent. Talent was never the issue.
My interest here, stated plainly. I co-founded KORE1 in 2005, the embedded software engineer staffing desk inside our engineering staffing agency practice earns a fee when a search closes through us, and I would rather you called. Nothing on this page requires it. Take the template, post it yourself, never fill out the form. That happens constantly and the page still did its job. Genuinely fine.
One thing to settle before the template. If the seat you are filling builds services, APIs, or product features that never touch a microcontroller, you are on the wrong page and this template will actively mislead you. Our software engineer job description template is built for that role. Use it instead.
The Posting Has to Say Where in the Stack This Person Lives
Before the posting, one thing has to be settled that the posting cannot settle for you. If you do not yet know whether this req is an embedded software engineer, a firmware engineer, or a systems-level architect, sort that first, because those are three different candidate pools and three different budgets. Our guide to hiring embedded systems engineers draws those title lines properly. Come back once the title is settled. We will wait.
Settled? Good. Now the posting has one job, and it is not listing languages.
Language filters almost nobody out here. Every candidate in this pool writes C. What separates them is the layer, meaning how far the person sits from the metal, because that distance decides what they debug at two in the morning and what tools they reach for when the board stops responding. Different distance, different hire.
The table below is not there to help you pick a title. It is there so you can find your row and copy what belongs in the posting. Bands are what our engineering desk has actually placed at over the past year, mid through senior. Not survey averages.
| Layer | What the person actually owns | Name this in the posting | Base band, mid to senior |
|---|---|---|---|
| Bare-metal firmware | Bootloader, HAL, device drivers, interrupt handling, memory-mapped I/O, power states | The MCU family and vendor toolchain. STM32 and Nordic nRF52 are not interchangeable experience. | $140,000 to $195,000 |
| RTOS application | Task design, scheduling, priority and timing budgets, inter-task communication, watchdog behavior | Which RTOS, and whether the person configures it or has modified the scheduler | $145,000 to $200,000 |
| Embedded Linux and BSP | Yocto or Buildroot layers, kernel configuration, device tree, driver authorship, board bring-up | The SoC, the build system by name, and whether they own the BSP or consume someone else’s | $150,000 to $205,000 |
| Certified and safety-critical | All of the above plus requirements traceability, structural coverage, and the audit evidence package | The standard and the level. Not “safety-critical experience preferred.” | $165,000 to $225,000 |
The bare-metal engineer and the embedded Linux engineer both put “C” and “Linux” on a resume. They are not close to the same hire. One has spent six years inside vendor HAL code and reads reference manuals for sport. The other has spent six years inside bitbake recipes and kernel configs and has not touched a register directly since school. Both wrote C.
Ask either one to do the other’s job and you get a competent engineer having a bad quarter. Sometimes two.
The posting is where you prevent that, and it takes one honest sentence. “You will own the Zephyr application layer on an nRF5340, working alongside the engineer who maintains our bootloader” tells a candidate more than fifteen bullets of cross-functional collaboration language ever will. Candidates self-select correctly when you give them enough to self-select against. Give them the layer.
Where the Hardware Sits Is a Term of Employment
Nobody writes this down. Every embedded search runs into it in week three.
Embedded work is coupled to a physical object. A real one. When the board misbehaves you need a scope on it, a logic analyzer on the bus, a debug probe attached, and frequently a hardware engineer standing next to you disagreeing about whose problem it is. Not a remote-friendly loop. Never has been. It is not remote-hostile either, and the honest answer for most teams sits somewhere in between, but the posting has to say where.
Four things a good embedded candidate wants to know before they apply, and almost no posting answers any of them:
- Do they get their own hardware, or is there one shared unit on a bench that three engineers sign up for?
- What is the lead time when a board dies? A week is normal. Six weeks happens, and it changes what the job feels like.
- Is there a remote debug rig, and does it actually work, or is it a wiki page from 2023 that nobody has tested?
- How many days on site? Say the number.
Industrial and controls work is the strict end of this. Our recruiters filter for candidates inside roughly 25 to 30 miles of the facility on those searches before anybody gets presented, because geography is not negotiable when the equipment is bolted to a plant floor. Consumer and IoT product teams have far more room. A Zephyr application engineer on a well-instrumented team can be genuinely productive three days a week from home. Bring-up weeks are still bring-up weeks. Everyone comes in.
Say which of those you are, in the posting, in plain language. Postings that stay vague on hardware access get applications from people who assume the answer they want, and you find out in the offer conversation. Expensive week to lose.
If the Product Is Certified, Name the Standard and the Level
“Safety-critical experience preferred” is the single most wasted line in embedded postings. It reads optional. In regulated work it is the entire job.
An engineer who has shipped under DO-178C has not merely written careful code. They have produced requirements traceability down to the source line, structural coverage evidence, tool qualification records, and a package that survived an auditor who was paid to disbelieve it. That is the job. That skill is a documentation and process discipline as much as an engineering one, and it does not transfer from a consumer product background in a quarter.
| Domain | Standard | The level your posting must state |
|---|---|---|
| Aerospace and avionics | DO-178C | Design Assurance Level A through E. DAL A carries the heaviest evidence burden. |
| Medical devices | IEC 62304 | Software safety class A, B, or C |
| Automotive | ISO 26262 | ASIL A through D |
| Industrial and process control | IEC 61508 | Safety Integrity Level 1 through 4 |
A DAL C avionics background and a DAL A avionics background are separated by years of process exposure, and candidates know exactly which one they have. Ask for the level and they tell you in the first paragraph of the cover letter. Ask for “safety-critical experience” and you get everybody who has ever worked near a medical device, and I do mean everybody.
While you are in there, check your coding standard reference. A great many postings still ask for MISRA C:2012 compliance. MISRA C:2025 has been the current guideline set since March 2025, and MISRA C:2023 sat in between. Small thing. Real damage. An engineer who works to these standards daily reads a 2012 reference and concludes your process has not been touched in over a decade, which may be unfair and is not something they will stay to investigate.
Embedded Software Engineer Job Description Template
Written for a mid to senior embedded software engineer on a product team. Bracketed text is intake guidance for you, not public posting copy. Replace every bracket before you post, and if you cannot answer one of them, that is the conversation to have with the hiring manager before the req opens.
Job Title: Embedded Software Engineer [or Firmware Engineer / Senior Embedded Software Engineer / Embedded Linux Engineer. Pick the one your candidates search for, which is usually the layer, not the seniority]
Location: [City, State. State the on-site days explicitly. “Hybrid” without a number reads as “on site five days” to an experienced candidate]
Employment Type: [Full-time / Contract / Contract-to-Hire]
Department: [Firmware / Embedded Systems / Platform Engineering / R&D]
Reports To: [Firmware Manager / Director of Engineering / VP of Hardware Engineering]
About the Role
We are hiring an embedded software engineer to own [specific scope: the application layer on our next-generation sensor platform / the bootloader and OTA update path across our product line / the Yocto BSP for our new gateway hardware]. You will work on [name the target: an ARM Cortex-M33 running Zephyr / an NXP i.MX 8M running embedded Linux / a Renesas RA family MCU on bare metal], alongside [number] hardware engineers and [number] other firmware engineers. The product [ships to X customers / is in its third revision / enters certification in Q2], and the code you write will be under support for [realistic number] years.
What You’ll Do
- Own [drivers, bootloader, RTOS task design, BSP layers. Name two or three areas, not eight] on [the target platform, named]
- Debug at the hardware boundary using JTAG or SWD, a logic analyzer, and an oscilloscope, working directly with hardware engineers through board bring-up and revision cycles
- Own the timing and memory budgets for your subsystem, including the unglamorous work of proving a change did not cost 3KB of RAM the product does not have
- Write and maintain on-target tests, plus whatever hardware-in-the-loop coverage the team runs, so regressions surface before a unit reaches a customer instead of after
- Participate in code review with the level of rigor a ten-year support window requires
- [If regulated: produce and maintain requirements traceability and verification evidence to (STANDARD, LEVEL), working with our quality team through audit cycles]
What We’re Looking For
- [3+ / 5+] years writing production embedded software that shipped on hardware customers actually received
- Depth in C, and [C++ if the codebase genuinely uses it. If your codebase is C99 and someone wrote “C/C++” out of habit, delete the C++]
- Hands-on experience with [name the specific MCU or SoC family. “ARM experience” covers a Cortex-M0 and a 64-bit Cortex-A78 and means nothing on its own]
- Comfort reading a schematic well enough to know whether the bug is yours or the hardware team’s, and the judgment to raise it either way
- [RTOS: name it. FreeRTOS, Zephyr, ThreadX, VxWorks, QNX. Say whether the person configures it or extends it]
- Working knowledge of [the peripherals and buses this role touches: I2C, SPI, UART, CAN, USB, BLE, Ethernet. List what is on your board, not the full catalog]
Nice to Have
- Experience through a product certification cycle [name the standard]
- Familiarity with our build and CI setup [Yocto, Buildroot, CMake, West, GitHub Actions with hardware runners]
- Exposure to secure boot, signed firmware images, or OTA update architecture
- Power profiling and battery life optimization on a constrained device
Hardware and Location
[This section is why the template exists. Fill it in.] You will have [your own dev kit and debug probe / access to a shared bench with X units]. Our lab is in [city], and this role is on site [number] days per week, with bring-up periods requiring more. Board turnaround when hardware fails is typically [X] weeks.
Compensation
[$X to $Y] base, plus [bonus and equity structure]. Contract rates for embedded engineers run $70 to $100 an hour at mid-level and $100 to $140 for senior bare-metal or safety-critical work. Benchmark your band against the salary section below, or run your specific role through the KORE1 salary benchmark tool before the req opens rather than after two candidates decline.
Embedded Software Engineer Salary in 2026
Compensation data for this title is unusually noisy, and the reason is worth understanding before you price a req against it.
| Source | Title measured | Reported figure, United States |
|---|---|---|
| ZipRecruiter (Aug 2026) | Embedded Software Engineer | $153,383 average. 25th percentile $131,500, 75th percentile $173,000 |
| Glassdoor (2026) | Embedded Software Engineer | $174,217 average total pay. Typical range $141,954 to $216,887 |
| Levels.fyi (2026) | Embedded Systems Software Engineer | $145,000 median. Company medians span $128,000 at Texas Instruments to $324,000 at Nvidia |
| PayScale (2026) | Embedded Software Engineer | $91,398 average base, ranging $70,000 to $126,000 |
That is an $83,000 gap between two averages for one title. It is not measurement error. It is signal.
Part of it is definition. Glassdoor reports a total pay estimate that folds in bonus and stock, PayScale reports base and skews toward self-reported early-career data, and Levels.fyi draws heavily from large public technology companies where equity does most of the work. Three different questions. Read each column for what it measures rather than averaging them together.
The rest of it is the title collision this whole page is about. A firmware engineer maintaining motor control code at an industrial manufacturer in Ohio and an embedded systems engineer writing GPU firmware at Nvidia both appear in these datasets under the same words. Same three words. The Levels.fyi spread between Texas Instruments and Nvidia, $128,000 against $324,000, is the clearest picture of it anyone publishes.
Price your req against your layer, your industry, and your metro. Then check it twice. KORE1 staffs across 30+ U.S. metros and the pattern holds consistently: an embedded seat in Austin, Phoenix, or the Raleigh-Durham corridor prices ten to twenty percent under Bay Area and Seattle at an equivalent technical bar. Certification exposure moves it the other direction regardless of geography, which is one of the few premiums that has not compressed since 2022. It has held.
For context on the broader market, the Bureau of Labor Statistics projects 15% employment growth for software developers, quality assurance analysts, and testers from 2024 to 2034, with roughly 129,200 openings a year. Embedded sits inside that number without being visible in it. The specialization is smaller, older on average, and considerably harder to hire into, which is the part the aggregate growth figure hides.
Questions That Come Up on Embedded Reqs
Our device runs embedded Linux. Is that still a firmware hire?
Different job, same title. Embedded Linux work centers on Yocto or Buildroot layers, kernel configuration, device tree, and driver authorship, which is a distinct skill set from bare-metal firmware. Some engineers do both. Most have a clear primary, and your posting should name which one you need. If the answer is genuinely both, say that too, and expect the search to take longer and cost more.
Should the posting name the specific MCU family?
Yes, and be more specific than the vendor. “STM32H7” tells a candidate something. “ARM experience” does not, because ARM spans a Cortex-M0 running on a coin cell and a 64-bit applications processor. Naming the part also signals that the req was written by somebody who knows the product, which is a stronger recruiting signal than any culture paragraph you could put above it.
How much of this job can happen away from the lab?
Two to three days a week from home is realistic for most application-layer embedded roles on a well-equipped team. Bring-up, hardware bug triage, and certification testing pull people back to the bench. Industrial and controls roles run tighter, often fully on site, because the equipment does not move. Whatever your real answer is, put the number in the posting. The number, not the word. Candidates who need a different arrangement will screen themselves out in the first thirty seconds instead of the fourth interview.
We put “safety-critical experience preferred” in the req. Does that filter anyone?
Almost nothing useful comes back from that line. It reads as optional to the engineers who have that background and as an invitation to the ones who do not. Name the standard and the level. “DO-178C DAL B” or “IEC 62304 Class C” is a phrase that a qualified candidate recognizes instantly and an unqualified one skips past, which is the entire function of a requirement.
What does certified experience add to the band?
$15,000 to $30,000 over an equivalent uncertified seat, based on what our engineering desk has placed at over the past year. Not for the coding. The premium buys traceability, verification evidence, and audit process, which take years of exposure to learn and cannot be picked up from documentation during onboarding.
How long do embedded searches actually run?
Longer than general software searches, and the gap is mostly logistics rather than talent supply. Logistics, mostly. Our average time-to-hire across IT and engineering roles is 17 days when the req is accurate and the interview loop moves. Embedded searches with a named layer, a stated on-site expectation, and a realistic band land near that. The ones that stall are usually stalling on geography or a comp number set from a generic software survey, and both of those are fixable before the req opens rather than after week eight.
The Three-Line Check
Read your draft once more and check three lines. Does it name the silicon? Does it say where the hardware is and how many days you expect somebody at it? If the product is certified, does it name the standard and the level?
Most embedded postings fail all three, which is genuinely good news, because it means fixing them is available to you this afternoon and to your competitors only if they think of it. Fix yours today.
If you would rather not run this one yourself, talk to a KORE1 recruiter and we will scope the req with you before it goes live. That practice has been placing firmware, controls, and embedded Linux engineers for over twenty years, with a 92% twelve-month retention rate on those placements. We staff embedded roles as direct hire when the product has a long tail and as contract when it is a bring-up sprint with an end date. If the seat is squarely bare-metal, our firmware engineer staffing desk runs those searches specifically.
