Post a backend role and you will have two hundred applications by Thursday. That is not the problem. The problem is that roughly none of the signals in those applications predict whether the person can ship, and the signals that used to work have quietly stopped working, because a polished portfolio and a clean take-home assignment now cost a candidate an afternoon instead of a fortnight.
So the question is not where developers are, since they are everywhere and easy to reach. The question is who has already done the screening, how deep that screening actually goes, and what you still have to verify yourself before someone touches your repository.
Key takeaways
- Demand is not softening. The US Bureau of Labor Statistics projects employment of software developers, QA analysts and testers to grow 15% between 2024 and 2034, much faster than the average occupation, with about 129,200 openings per year across the decade.
- "Vetted" is not a regulated term. It ranges from an automated skills quiz to a multi-stage technical interview with reference checks, and the platforms using the word rarely tell you which end they sit at.
- Two commonly recommended sources no longer exist. GitHub's job board closed in 2021 and Stack Overflow Jobs was discontinued on March 31, 2022. Any 2026 guide still listing them was not checked.
- Take-home assignments have lost most of their signal value, so the weight has moved back to live conversation, code review discussion and paid trial work.
- Every platform claim in this category is a marketing claim until you see the artifact behind it. The verification table below tells you which artifact to ask for.
What does "vetted" actually mean?
A vetted developer is one who has passed structured screening beyond a resume review, typically covering technical assessment against realistic problems, communication evaluation, verification of past delivery, and in stronger processes a paid trial period before a long-term commitment.
The word does a lot of work and hides a lot of variance. Treat it as a spectrum with four rungs, and ask any provider which rung they are on before you accept the label.
| Rung | What it involves | What it actually proves |
|---|---|---|
| 1. Profile screening | Resume review, automated skills quiz, identity check | The person exists and claims the skill |
| 2. Technical assessment | Timed coding challenge or take-home task | They can complete an exercise, possibly with assistance |
| 3. Human technical interview | Live problem solving, system design, code walkthrough | They can reason out loud and defend decisions |
| 4. Verified delivery history | Reference checks with past clients, evidence of shipped work, trial period | They have actually delivered under real conditions |
Most platforms operate at rungs one and two while marketing at rung four. That gap is where the hiring risk lives, and closing it is your job rather than theirs.
Where can you find vetted software developers?
There are four distinct channels, and they suit genuinely different situations. Choose the channel before choosing the vendor.
| Channel | What you get | Screening depth | Best when |
|---|---|---|---|
| Curated talent networks | Individual contractors pre-screened by the platform | Usually rungs 2 to 3 | You need one or two specialists quickly and have technical leadership in house |
| Outsourcing and matching partners | Established providers and teams, not individuals | Varies widely, ask | You need a team, a delivery process, or continuity beyond one person |
| Open marketplaces | Large pools with ratings and work history | Rung 1, you do the rest | Small or short scopes, and you are willing to screen yourself |
| Direct sourcing | Candidates found through communities, code hosts and professional networks | None, entirely yours | You have strong engineering leadership and time to run a full process |
Curated talent networks
These platforms screen individuals and present a shortlist. Toptal built its brand on selectivity and states that only a small percentage of applicants enter its network. Turing markets AI-driven matching with an emphasis on timezone overlap and offers a trial window. Revelo focuses on Latin American talent for US buyers, which puts most of the working day inside the same timezone band.
Every sentence in that paragraph is a company claim, including the selectivity percentages you will see quoted everywhere. None of it is independently audited. Treat the claims as a starting point, then ask for the artifacts in the verification table below.
Outsourcing and matching partners
This channel supplies teams and providers rather than individuals, which matters when you need delivery continuity rather than a pair of hands. The tradeoff is that you are evaluating an organization, including its process maturity, security posture and staff retention, not just a person.
Vetting an organization is a different exercise with a different checklist, and the operational side of it, meaning whether a team can actually release and support what it builds, is covered in our production readiness checklist.
Open marketplaces
Upwork and similar platforms give you scale, ratings and work history. What they do not give you is screening. Ratings inflate over time, categories are gamed, and the profile that looks strongest is often the one optimized hardest for the platform's ranking rather than for your problem.
They work well for small, well-specified scopes where a bad outcome costs you a week. They work poorly for anything load-bearing.
Direct sourcing
Code hosts and professional networks remain excellent for finding people and useless for screening them. Public repositories show real work, real commit history and real code review behaviour, which no resume can fake, but reading that evidence requires engineering judgment you must supply yourself.
Two clarifications worth making, because outdated guides keep repeating them. GitHub's dedicated job board closed in 2021, so GitHub is a sourcing and evidence tool now, not a hiring platform. Stack Overflow shut down Jobs and Developer Story on March 31, 2022, stating that talent acquisition was not a space where it held a competitive advantage.
How do you verify a platform's vetting claim?
Ask for the artifact, not the adjective. This table is the shortest path from marketing language to evidence.
| What you are told | What to ask for |
|---|---|
| "Top 3% of talent" | The rejection criteria and what stage most candidates fail at |
| "Rigorous technical screening" | A description of the actual assessment, and whether a human reviewed it |
| "AI-powered matching" | Which signals are machine-ranked and who performs human validation |
| "Fully vetted" | Which of the four rungs above their process reaches |
| "Communication assessed" | Whether it was a live conversation or a written test |
| "Trial period available" | Length, cost, and what happens to the work if you decline |
| "Replacement guarantee" | Response window in business days, and any fee attached |
| "Timezone aligned" | Contracted overlap hours, not a country name |
The single most revealing question is which stage most candidates fail at. A platform with a real process can answer immediately and specifically. A platform with a light process will answer in adjectives.
What should your own screening look like?
Platform screening is a filter, not a substitute for your judgment. Five stages, in this order.
- Evidence review. Public code, shipped products, or a redacted work sample. You are looking for structure and consistency, not cleverness.
- Live technical conversation. Forty-five minutes on a real problem from your domain, with follow-up questions. This is where seniority actually shows.
- Code review discussion. Give them a small piece of imperfect code and ask what they would change and why. Reviewing well is harder to fake than writing well.
- Reference check. Two references you selected, not two they nominated, with one specific question: would you hire this person again for the same role?
- Paid trial. One to two weeks of real work on your process, with success criteria written before it starts.
Tip: ask every candidate to describe something they shipped that did not work and what they changed afterwards. Candidates who cannot answer this have either not shipped much or are managing you.
How has AI changed developer screening?
The take-home assignment is the casualty. An assistant will complete a standard four-hour exercise cleanly, which means a strong submission now tells you a candidate has access to tooling rather than that they can build. Grading them the way you did in 2022 measures almost nothing.
That does not mean disqualifying AI use, since practically every working developer uses assistants and the ones who do not are often slower for no benefit. What separates candidates now is judgment about the output: what they rejected, what they rewrote, and whether they can explain code they submitted line by line.
Three screening adjustments follow from this. Move weight from asynchronous exercises to live conversation, ask candidates to critique code rather than only produce it, and use the paid trial as the real evaluation, because a two-week engagement on your actual codebase cannot be shortcut by tooling.
Ask directly about tooling rather than banning it. "Which assistants do you use, and where do you not trust them" is a better seniority signal than any algorithm question, and the answer separates people with real production experience from people with fluent output.
What are the red flags when hiring a developer?
- A portfolio with no specifics. Projects described without stack, scale, role or outcome are usually borrowed.
- Reluctance to talk live. Anyone unwilling to hold a technical conversation is protecting something.
- Claimed expertise across everything. Fifteen languages at expert level means none of them.
- No verifiable history. Real developers have past clients, shipped products, or public work.
- Resistance to a paid trial. The trial is favourable to a good candidate, so refusal is informative.
- Communication that is already slow. Responsiveness before the contract is the ceiling, not the floor.
- A platform that will not describe its process. Vagueness about screening is a statement about screening.
How much do vetted developers cost?
Rates in this market are quoted, not published, and every figure circulating online traces back to vendor marketing rather than to a survey anyone can inspect. Rather than repeat numbers I cannot verify, here is the structure that determines what you pay.
Cost is driven by four things: seniority, geography, engagement length and who carries the employment overhead. Curated networks price above open marketplaces because screening costs money, and outsourcing partners price differently again because you are buying process and continuity rather than hours.
Compare on total cost rather than on rate. A cheaper engineer who needs more review, more clarification and more rework can cost more per delivered feature than a more expensive one, and that difference is invisible on a rate card.
Which channel you are pricing changes the calculation entirely, and the breakdown by service type is in our overview of development outsourcing services.
What mistakes do buyers make when hiring vetted developers?
- Trusting the word "vetted" without asking which rung. The term is unregulated and used loosely.
- Skipping your own technical interview. No platform knows your codebase or your standards.
- Using nominated references. Selection bias in the only independent evidence available.
- Grading take-home assignments the old way. That signal has degraded and it will not come back.
- Hiring the strongest interviewer. Interview performance and delivery performance are different skills.
- Committing before a paid trial. The trial is the cheapest information you will ever buy.
Get matched instead of screening from scratch
VettedOutsource matches companies with pre-screened development partners and engineers, so you start from a filtered shortlist rather than an open market. You still run your own technical interview, which you should, but you run it on three relevant candidates instead of thirty applicants.
