After three months, raising the budget feels like the obvious next step. Sometimes it is. But without knowing why relevant engineers drop out, a larger budget can simply fund another quarter of the same problem.
Check which job you are actually buying
FPGA and ASIC are not interchangeable search terms. A job description that covers RTL, verification, physical design, firmware and customer contact can hide three or four different roles. Even a very good engineer rarely has real depth in all of them.
Go back to the work that has to get done. What should be working after the first six months? Which stage of the project does this person take over? Which skill is missing from the team today, to the point that work stops without it? Derive the requirements from that, not the other way round.
Market mapping should test that defined profile. A list of people with the right term somewhere on a public profile is not evidence of available capability.
An industry shortage does not diagnose your search
The 2023 SIA and Oxford Economics report projected that roughly 67,000 new US semiconductor jobs could remain unfilled by 2030. That is a US projection, not a count of current Polish FPGA vacancies.
That context is a reason to treat access to these engineers seriously. It is not proof that the only fix for your search is a raise. A raise is the shortcut, and it leaves every other cause untouched. The answer has to come from the specific market (each Polish city is its own microclimate) and from specific conversations with candidates.
For the outreach challenge itself, read about reaching passive candidates who do not respond to job ads.
Separate four different failures
Capability: people respond but technical assessment shows they have not done the required work. Revisit search criteria and calibrate with the technical lead.
Access: the same profiles keep arriving and nobody has mapped teams working on comparable products. Expand the market map rather than searching the same database again.
Proposition: relevant people understand the opportunity but decline because of compensation, location, scope or technology. Here, changing the offer may be necessary.
Speed: candidates pass assessment and then wait between meetings or for a decision. Higher compensation does not remove a week-long feedback queue.
These failures can coexist. Keep them separate in the reporting so that one convenient explanation does not hide the others.
Has your FPGA or ASIC search stalled?
In 30 minutes, we can examine the brief, results and decline reasons to identify what actually needs to change.
Book a 30-minute callReconstruct the last 90 days
Collect the basics: number of unique people contacted, confirmed skills, replies, interviews, reasons for declining, assessment results and waiting time between stages. Separate no reply from a refusal (this matters more than it looks) and lack of experience from lack of interest.
Do not count the same person twice because two partners presented them. Do not treat a profile with no conversation as confirmed availability. If you cannot reconstruct this history, you do not yet have the grounds to judge which channels work.
If the budget rose by 15% tomorrow, which specific people could return to the process, and why? If nobody can answer, investigate decline reasons before changing the number.
Test one meaningful change
If office presence is the blocker, check whether it can be reduced where the work allows it. We have started many projects with "that is impossible in our organisation" and watched it turn into "worth trying". If lab access is required, describe the real working rhythm instead of writing "hybrid".
If the profile combines two specialisations, decide which one the person must bring on day one and which the team can cover. If pay is the issue, verify it against candidates with comparable experience. Do not change five conditions at once, or you will not see which one helped.
If the same unsuitable profiles keep returning, it is worth checking whether application volume is hiding a definition problem.
Two questions worth answering
Do we have to open up to other countries? Sometimes. First check the collaboration model, the presence requirements and what the project actually demands. A wider search area without a real ability to hire only gives you a bigger spreadsheet.
Does 90 days mean the partner failed? The number of days alone does not settle that. Having no knowledge after 90 days is the much worse signal. By then you should know the constraint and have a decision to make. A good partner does not just report back; they propose fixes along the way so the process keeps its momentum.