Frontend engineering

Hire frontend developers whose interfaces hold up on real devices

You know who is writing your interface code before they start. Say what the interface has to do; a shortlist usually follows within a few days.

Start with a shortlist

Tell us what the interface has to do, which devices and browsers matter, and the working hours you need us reachable in.

How we handle what you send is set out in our privacy notice.

Frontend work fails quietly: the build passes, the test suite is green, and the interface is still broken for a real person on a real device.

How we know

Three public accounts from working software engineers: a product completely broken in Safari with no acknowledgement a customer could find; the reasoning that keeps it shipped, "I wouldn’t say 4% is a meaningful market share, so ... leaving the sites broken in Safari"; and five repositories, six languages and three frameworks inherited undocumented from an outsourced contractor.

None of that shows up in a test report. The risk is not whether someone can write the code. It is that nobody on the project can see when it is wrong, and the person who finds out is a customer on a phone who never files a bug report.

Show data table
Industry data, not Atyantik data. WebAIM scanned one home page per site across the top one million sites with the WAVE engine, February 2026. Each bar is the share of home pages where that failure type was detected at least once, which measures detectable failures on home pages rather than sites that fail WCAG. Automated detection finds only a subset, so the true rate is higher.
Item Share of home pages with at least one detected instance
Low contrast text 83.9%
Missing image alt text 53.1%
Missing form input labels 51%
Empty links 46.3%
Empty buttons 30.6%
Missing document language 13.5%

All six are decided in frontend code, and together they account for 96 percent of every error the scan detected.

What an automated scan finds on the top one million home pages: WebAIM Million, February 2026 Industry data, not Atyantik data. WebAIM scanned one home page per site across the top one million sites with the WAVE engine, February 2026. Each bar is the share of home pages where that failure type was detected at least once, which measures detectable failures on home pages rather than sites that fail WCAG. Automated detection finds only a subset, so the true rate is higher. WebAIM Million 2026, WebAIM, Institute for Disability Research, Policy, and Practice, Utah State University, https://webaim.org/projects/million/

These have been the six most common for seven years running, and you can run the same check against your own site today without involving us. The standard is WCAG 2.2 Level AA, published by the W3C Accessibility Guidelines Working Group on 12 December 2024. Separately, the European Accessibility Act has applied since 28 June 2025 and covers e-commerce; it routes conformity through EN 301 549 rather than naming a WCAG level. Of ten of the pages ranking for this work that we read, none names a WCAG version or level as a deliverable.

Show data table
Industry data, not Atyantik data. Reported by the HTTP Archive Web Almanac 2025 Performance chapter from July 2025 Chrome UX Report data. Mobile and desktop are separate CrUX origin sets over overlapping but not identical populations, so the eight-point gap on all three is a difference between two groups of tracked origins, not one site measured twice.
Dimension Mobile Desktop
Largest contentful paint 62% 74%
Interaction to next paint 77% 97%
Cumulative layout shift 81% 72%
All three 48% 56%

Interactivity on desktop is close to saturated at 97 percent. The deficit sits on phones, and most of it is loading.

Origins tracked by the Chrome User Experience Report that pass each Core Web Vital, July 2025 Industry data, not Atyantik data. Reported by the HTTP Archive Web Almanac 2025 Performance chapter from July 2025 Chrome UX Report data. Mobile and desktop are separate CrUX origin sets over overlapping but not identical populations, so the eight-point gap on all three is a difference between two groups of tracked origins, not one site measured twice. HTTP Archive Web Almanac 2025, Performance chapter, https://almanac.httparchive.org/en/2025/performance

Fast is a threshold with a number and an issuer behind it: largest contentful paint at 2.5 seconds or less, interaction to next paint at 200 milliseconds or less, cumulative layout shift at 0.1 or less, all at the 75th percentile of real loads. Those come from the Chrome Web Vitals initiative, and interaction to next paint replaced first input delay in March 2024. Ask any candidate what their last project scored on the three. If the problem is speed rather than headcount, making an existing page fast is a different engagement.

What has to pass before your main branch

  1. Review

    A senior software engineer reviews every change before it merges.

    Not a rubber stamp on a green pipeline. GitFlow, unit and integration tests, CI/CD gating every merge, and architecture decisions written into the technical requirements specification as they are taken. A regression stops the merge rather than accumulating until someone complains. A pull request that will not merge without a named reviewer, a passing suite and a linked decision record.

  2. Contrast and labels

    Frontends meet WCAG 2.2 by default, and higher levels are supported on request.

    That is the deliverable, not a report about it, and it is the default rather than something you fund separately. The six failure types above, addressed at build time rather than by an audit six months later.

  3. Browsers and devices

    Which browsers and devices have to work is agreed at the start, not assumed.

    A WebKit failure goes unnoticed longest, because the people building an interface are rarely the people using that browser. So the supported set is written down before the work starts, rather than discovered by a customer. A named list of browsers and devices in the engagement, so nobody argues later about whether Safari counted.

  4. Budgets

    Frontends meet performance budgets on every device. Speed is a quality criterion, not a feature.

    Where the three metrics above are in scope, they get numbers at the start, agreed with you, and a regression stops the merge. A Lighthouse budget file in the repository with the three thresholds, versioned alongside the code.

Who reviews the assigned developer's code, and against what standard, is rarely written down anywhere you can check before you sign. Ours is above. Hold us to it, and ask whoever else is bidding for theirs in the same form.

50+Engagements delivered across 7 countries

Atyantik metrics, verified 2026-05-14

3.5 weeksAverage MVP launch

Atyantik metrics, verified 2026-05-14

3 weeksPilot against criteria agreed in writing first

Atyantik engagement terms, verified 2026-05-14

Bob Miller, President of FoxDealer, put his name to this in 2022: "Incredible partners who delivered a critical project on an abridged timeline. The team blew everyone away with what we accomplished." Every other client engagement is described without names, by class rather than by company.

If it goes wrong

  • You interview one person and somebody else turns up on day one.

    One practitioner account records asking for a senior in one stack and being given a junior in another, because that was who was on the bench.

    What we commit to

    You interview the person proposed, and that is the person assigned. If they become unavailable, you are told before it happens and you interview the replacement on the same terms. Everyone we propose is checked by our own technical team first, and if our bench does not match what you need we say so and recruit.

  • They are not good enough and you are already a month in.

    The criteria are agreed in writing, so the bar is not renegotiated once the work is seen.

    What we commit to

    The first two to three weeks are a pilot against evaluation criteria written into the engagement before anyone starts. Miss them and we replace the developer and re-run the pilot against the same criteria.

  • Your security review arrives and the whole thing goes quiet for two weeks.

    What we commit to

    NDA from day one, single sign-on, your secure repositories, agreed working hours. Code lives in your repositories and deployments in your cloud accounts from the first commit, so there is no environment of ours to assess.

  • You want to end it one day and find you cannot.

    What we commit to

    You own every line from day one. On offboarding you keep the code, the operational runbook and the documentation, and the handover runs as a written process rather than a standoff.

  • We are the wrong team for the work and nobody says so until it is expensive.

    What we commit to

    If the fit is not right, we tell you. We have declined work that would have suited us and not the project.

Whether the person you interview is the person you get, what you receive on the last day, and how a security questionnaire gets answered are the three things that usually go unwritten. All three are above. Take these five commitments to whoever else is bidding and ask for their version in writing.

Start the pilot, or take the shortlist

Send what the interface has to do and you get names. A senior software engineer replies, usually within 24 hours.

How we handle what you send is set out in our privacy notice.

No number comes back from this form, because a form cannot produce an honest one. What comes back is a person and a conversation. If you would rather size the cost yourself first, keep reading.

Show data table
Industry data, not Atyantik data, and United States only. The salary line is the Stack Overflow 2025 Developer Survey median for front-end developers, self-reported by 23,928 salary respondents rather than measured by a government agency. Payroll taxes are IRS 2026 statutory rates. Cost-per-hire is the SHRM benchmark, a fiscal year 2015 figure pooled across nonexecutive roles and nine years older than the salary it is added to. A floor, not an estimate: four further components are excluded.
Step Change Running total
Median salary, self-reported 145,000 USD 145,000 USD
Employer FICA at 7.65% 11,092.5 USD 156,092.5 USD
Net FUTA 42 USD 156,134.5 USD
Average cost-per-hire 4,129 USD 160,263.5 USD
First-year floor 160,263.5 USD

About 160,000 USD before benefits, equipment, or an agency placement fee.

What a United States in-house frontend hire costs in year one, at floor Industry data, not Atyantik data, and United States only. The salary line is the Stack Overflow 2025 Developer Survey median for front-end developers, self-reported by 23,928 salary respondents rather than measured by a government agency. Payroll taxes are IRS 2026 statutory rates. Cost-per-hire is the SHRM benchmark, a fiscal year 2015 figure pooled across nonexecutive roles and nine years older than the salary it is added to. A floor, not an estimate: four further components are excluded. Stack Overflow 2025 Developer Survey; IRS 2026 statutory FICA and FUTA rates; SHRM Human Capital Benchmarking cost-per-hire, fiscal year 2015

Four things are left off: employer health and retirement contributions, agency placement fees, equipment and software, and the ramp before a new hire is productive. All four are real, none could be sourced to a named publisher for this role, so none is drawn. Put your own numbers against them and the total only goes up. In-house is often still the right decision when the work is permanent. What it is not is a salary line, and a salary line is the comparison that makes every outside team look expensive.

Three shapes an engagement takes

  • Pilot

    3 weeks

    What it includes

    • Evaluation criteria agreed in writing before the first day
    • A real piece of your backlog rather than a test task
    • The merge gate running from day one
    • A written assessment at the end whether you continue or not

    Widens when The codebase you hand over is undocumented, so the first days go on reading rather than shipping

  • Full-time dedicated software engineer

    3 months by default, then rolling

    What it includes

    • One named software engineer in your repositories
    • A daily standup held in your working hours
    • Senior code review on every change
    • DevOps and QA added if the scope needs them

    Widens when You need more than one person, a higher seniority mix, or DevOps and QA inside the engagement rather than beside it

  • Part-time or specialist

    Scoped to the work

    What it includes

    • A defined piece of work with its own acceptance criteria: an accessibility remediation, a performance budget brought back under threshold, or a design system built out

    Widens when Conformance to a named accessibility level is in scope rather than best effort, which changes what has to be tested and evidenced

Six things move the number and every one is your decision. How many people and at what seniority mix. Whether DevOps and QA sit inside the engagement or beside it. Which of your working hours the team has to be reachable in, beyond the daily standup we hold in them. Whether we inherit an existing codebase or start clean. Whether conformance to a named accessibility level is in scope or best effort. And how long it runs.

We publish no rate, because the honest one depends on those six. What we will not do is promise a quote a form cannot produce.

Before you send anything

These come up in every scoping call, so here they are upfront.

Can we interview candidates?
Yes, if you want to. Some clients run their full hiring loop and others skip interviews entirely. If a shortlist does not produce the right fit, we send another.
Will they work in our timezone?
Our people are currently all based in India and work IST 10:00 to 19:00. Rather than quote an overlap window, we hold a daily standup inside your working hours: around thirty minutes with the lead, and the team when the work needs them.
What if the fit is not right?
We replace the developer and re-run the pilot against the same criteria, which were agreed in writing before anyone started.
Who owns the code, the cloud and the data?
You do. From day one every line is your IP, and any documentation we produce is yours to keep. If you do not have your own hosting and accounts, we provision them and hand over the keys.
What if we stop working together?
Every engagement starts with documentation: architecture, deployment, pipelines and change notes. If you move the work in-house or to another partner, the handoff is a written process and you keep the operational runbook.
How do you keep the code solid rather than duct-taped together?
The merge gate above is the answer, and it runs from day one rather than at review time. Architecture decisions go into the technical requirements specification, so anyone reading the system later knows why it is shaped that way.
Do you support DevOps and QA too, and is there a minimum commitment?
Yes to the first: full-team buildouts pair developers with at least one DevOps and one QA engineer, and if your engagement only needs developers we scope around that. On the second, most engagements run three months or longer, because that is when a new developer’s productivity curve flattens. Shorter is possible for a scoped pilot or specialist project, and the length follows the work rather than the contract.

When we are the wrong call

  • You have an interface that has to work for real people on real devices and nobody who owns that.

  • You are adding frontend capacity to a team that already exists and want people your own engineers can review.

  • You are inheriting a codebase somebody else wrote.

  • You want the engagement to be able to end cleanly, in writing.

Where to go instead

Tell us what the interface has to do

A shortlist usually follows within a few days, and you can interview them or skip it.

Whether the interface exists and is failing on phones or does not exist yet, the first conversation is the same: what it has to do, which devices and browsers matter, and what you need to be holding at the end. You will speak to a senior software engineer, not an account manager working from a script. If we are not the right team for it, we will say so.

No pressure. No vague proposals. Just clarity.

Tirth Bodawala, Chief Technology Officer of Atyantik Technologies
Tirth BodawalaChief Technology Officer, Atyantik Technologies
  • We usually reply within 24 hours
  • A technical person answers, not a salesperson
  • Everything you share stays private
  • Pilot criteria agreed in writing before anyone starts

Get in touch

A senior software engineer reads this and replies, usually within 24 hours.

How we handle what you send is set out in our privacy notice.