9 Checks That Finally Found Us a Drone Software Engineer in Dubai After 2 Failed Searches

The first search ran eleven weeks and ended with a declined offer. The second ran nine and ended with a hire who left in month five, perfectly competently, because we had hired a simulation specialist for a job that needed someone who had watched their own code misbehave two hundred feet up. The third search took five weeks and worked. What changed was not the advert or the salary. It was nine checks, in order, and a ninety-minute exercise that I now refuse to skip.

Why This Role Is Harder to Hire Than Its Salary Suggests

Dubai has a visible reason to need these people. Joby holds a six-year exclusive agreement with Dubai’s Roads and Transport Authority to operate air taxi services, with four initial vertiports planned at Dubai International, Palm Jumeirah, Dubai Marina and Downtown, and reporting in 2026 puts the start of operations at the end of this year. Around that sits a wider set of uncrewed work: inspection of towers and pipelines, survey and mapping, security, last-mile delivery pilots, and the fleet software that any of those needs once you operate more than three aircraft.

So demand is real and local. The problem is that the job title has been borrowed by four distinct engineering disciplines, and the market has not sorted them out. When you post for a drone software engineer in this market you will receive applications from all four, plus a tail of hobbyist pilots with Python and a tail of web developers who have built a dashboard that displayed a drone’s position. Reading that pile against a single scorecard is how both of my first two searches went wrong.

The Nine Checks, in the Order I Run Them

1. Decide which of the four jobs you are hiring, before you write a word of the advert.Write one sentence naming the deliverable in six months. “The aircraft holds position in a 25 knot crosswind” is a flight stack role. “One operator supervises eight aircraft” is fleet operations. If you cannot write that sentence, you are not ready to open the role, and no amount of screening will rescue it later. When we did this properly on the third attempt, two of our four shortlisted candidates from search two turned out to have been strong people in the wrong column.
2. Check flown hours against simulator hours.Ask directly: how many hours has your own code spent in the air, against how many in simulation, and what broke the first time it left the simulator. The ratio matters less than the quality of the answer to the second half. Everyone who has flown their own software has a story about the thing the simulator did not model, and they tell it with specifics: a vibration mode, a magnetometer reading corrupted by current draw, a timing assumption that only held on the bench. Candidates whose experience is simulation-only answer this question in generalities. This is the one place where the domain really cannot be taught quickly, and it is why I want at least one person on the team who passes it unambiguously.
3. Check the failure taxonomy.Ask the candidate to list the failure modes they have personally handled, ranked by how they would prioritise them. Strong answers lead with loss of link and sensor disagreement, and they talk about what the aircraft should do rather than what the alert should say. Weak answers lead with battery level, because that is the failure mode that is visible from a dashboard. There is no trick here; the ranking is the signal.
4. Check for real-time discipline.Scheduling, jitter budgets, watchdog behaviour, and what happens when a control loop misses its deadline. I ask: your loop runs at 400 Hz and one iteration takes 4 milliseconds, what do you do? The answer I want involves knowing that the question is about the next iteration and the one after, not about optimising the slow function. Engineers from web and mobile backgrounds, however senior, often do not have this reflex, and it is the most common reason a technically excellent candidate is wrong for a flight stack role. Our guide to hiring embedded and firmware engineers in Dubai goes deeper on testing the same reflex in a non-aviation context.
5. Check the approvals literacy, not the certificates.You are not hiring a compliance officer, and a pilot licence is not a software qualification. What you need is someone who understands that an airspace approval is a constraint that has to be expressed in software: a geofence that cannot be overridden by a mode change, an altitude ceiling that survives a mission replan, an audit trail that a reviewer can read. Ask how they would make a restriction impossible to violate rather than merely monitored. Candidates who have operated under approvals answer structurally; candidates who have not answer with warnings and pop-ups.
6. Check how they test what they cannot fly.Three techniques matter: hardware in the loop, replay of recorded flight logs against new code, and deliberate fault injection. Ask which of the three they have built rather than used. The person who has built a log replay harness has saved an employer a great deal of money and will do it again in your first quarter. This question also surfaces the candidate’s attitude to flight test time, which in this market is scarce and expensive: good engineers treat it as the most limited resource in the programme and design to use less of it.
7. Check the heat and dust answers.Gulf conditions are a genuine engineering problem and the answers are observable. Thermal derating on sustained summer operations, dust ingress and its effect on optical sensors, satellite navigation multipath between tall towers, and wind gradients over water near the coast. A candidate who has worked in the region or in a comparable climate will raise at least two of these unprompted. One who has worked only in temperate conditions will not, which is not disqualifying, but it tells you what the first three months of ramp-up will involve and lets you plan for it instead of discovering it.
8. Run the log-reading exercise.This is the check that changed our hit rate. Take a real flight log from your own operation, anonymise it, pick one with an anomaly that a careful reader can find, and give the candidate ninety minutes with the tooling of their choice to tell you what happened. It takes one of your engineers about four hours to prepare the first time and it is reusable. What it measures is exactly the job: patience with data, willingness to say “I do not know yet, here is what would settle it”, and the instinct to look at the time axis before the values. In three searches it has been a better predictor than every interview question I ask, including the ones above, and it is the only stage I would refuse to drop under time pressure. For the broader design of exercises like this, we set out the approach in hiring robotics engineers in Dubai.
9. Check the landing conditions before the offer goes out.Three questions, answered in writing, before anyone sees a number: when can this person actually start given relocation, what hardware will be on their desk in week one, and who owns the airframe they will test on. My first search collapsed at the final stage because the candidate asked the third question and we did not have an answer. It was an entirely reasonable thing to ask and an entirely avoidable loss. Engineers in this field are choosing between employers on whether the work will reach the air, and a vague answer here reads as a programme that is not real.

How the Nine Checks Fit Into a Five-Week Process

The order is not decorative. Checks 1, 2 and 3 are cheap and they eliminate most of the pile, so they go first. Check 8 is expensive for both sides, so it sits late, after you are confident enough to spend a candidate’s evening on it. Check 9 happens before the offer rather than after, which is the only change I would insist on if someone adopted nothing else from this article.

Where to Actually Look in This Market

For fleet operations and ground software, the productive adjacent pools in the UAE are logistics and dispatch platforms, fleet telematics, and industrial IoT. People who have built a system that assigns vehicles to jobs in real time and copes with vehicles disappearing already understand most of the problem; the aircraft is a detail. If you want the local landscape for that adjacency, our overview of Dubai logistics technology maps where those teams sit.

For perception, the adjacent pool is anyone who has shipped computer vision against a moving camera in uncontrolled lighting: automotive, retail analytics, inspection. The gap to close is calibration and timing rather than models. Our guide to hiring a computer vision engineer in Dubai covers the screening for that pool in detail, and almost all of it transfers.

For flight stack, be honest with yourself that you are running an international search for the senior hire, and use the programme as the pitch rather than the package. The people who are good at this are not short of offers, and the thing that moves them is whether their code will fly, how soon, and how often. Dubai can answer all three better than most places right now, and that is a stronger message than a number.

Three Things I Would Not Do Again

I would not screen on tool names. Filtering on a specific autopilot codebase or a specific middleware feels rigorous and is mostly a proxy for which employer someone happened to work for. The transferable part is the reasoning about timing, failure and test, and that is what checks 2 to 6 are for.

I would not put the hardest exercise first to save time. I tried it once, reasoning that the log exercise was so predictive that it should go early. Strong candidates declined to spend ninety minutes on a first contact, which is a reasonable position, and the people who accepted were disproportionately the ones with the least going on. Expensive stages have to be earned on both sides.

I would not hire two of the four columns into one person on purpose. People who genuinely span flight stack and fleet operations exist, and they are wonderful, and building a plan around finding one is how an eleven-week search happens. Hire two people, or hire one and name the gap.

FAQ: Hiring Drone Software Engineers in Dubai

Sources: Joby Aviation press release, 11 February 2024, for the definitive agreement with Dubai’s Roads and Transport Authority, the six-year exclusive right to operate air taxi services, and the four initial vertiport sites at Dubai International, Palm Jumeirah, Dubai Marina and Downtown. Time Out Dubai, 2026, for the reported expectation that air taxi operations begin by the end of 2026. Launch timings in this programme have moved more than once and the dates above are as reported rather than confirmed outcomes. The four-column split, the nine checks, the log-reading exercise, the funnel figures and the timing estimates are our own, drawn from three searches we ran for this profile in the UAE, and are not claims made by any company named above.