How a DevTools team sourced to hire backend engineers in 10 weeks

To hire backend engineers who understand distributed systems, a remote US DevTools company stopped waiting on job posts and sourced everyone in The Cognitive. Customer-reported: 420 engineers contacted, 31% took an AI interview, 5 were hired in 10 weeks, and search to signed offer took 19 days. None came from job boards.

What did the DevTools company's backend hiring look like after?

The first 4 rows are the customer's own figures. The last 2 are our arithmetic on them. The before column is mostly empty because nothing was sourced before.

Customer figures from our case study record dated August 2026. The 6+ weeks was an agency's quoted lead time, not a measured result.

Why couldn't the DevTools company hire backend engineers from job posts?

The customer is a DevTools SaaS company, remote across the US, that needed backend engineers who understand distributed systems. Its job posts drew plenty of applicants. By the team's own count, about 90% of that inbound pipeline did not match what the roles needed.

The reason is simple. Engineers with real distributed systems depth are usually employed, well paid and not reading job boards. Filtering inbound harder could not fix a pipeline that rarely contained the right people. The roles could only be filled by finding engineers who were not looking and giving them a reason to talk.

The market backs this up. The BLS puts median pay for software developers at $135,980 in May 2025 and projects about 106,100 openings a year for the group (BLS Occupational Outlook Handbook, read 5 October 2026). Ashby's 2023 sourcing data put engineering reply rates at 21%, among the lowest it measured (Ashby, read 5 October 2026).

The candidates we actually wanted were never going to apply. Sourcing plus a real technical interview in one pipeline is the whole game. Head of Engineering, DevTools SaaS Company, Remote US

What did a backend engineer recruiting agency quote?

Recruiting agencies quoted $25,000 per placement with a 6-week lead time. For 5 hires, that is $125,000 in placement fees before anyone wrote a line of code. The lead time was the bigger problem: the record says the roadmap could not absorb 6 weeks of waiting before candidates even arrived.

So the Head of Engineering's team took the work in house. The plan was to build the pipeline rather than filter one: write what the role really needed, find the people who matched it, reach them on channels they answer, and give every interested engineer a technical first round without pulling a senior engineer into a screening call.

Set against pay, $25,000 is 18.4% of the BLS median for software developers, on our arithmetic. The 6 weeks were the harder cost to carry. While a backend seat sat empty, roadmap work waited on the engineers already on the team, and the agency's clock ran before a single candidate arrived.

How did they source and interview backend engineers step by step?

The record describes the rollout. Each step is written in today's product terms, because the setup dates from mid 2026 and the product has changed since.

1. Turn the engineering brief into plain English Write the role the way the team talks about it, with the distributed systems experience as a must-have. The brief is read into editable filters across ~900M public profiles. Example brief, ours: "Backend engineer who has run services in production and owned a distributed data store or queue, Remote US."
2. Test the first long-list against the team The record says the long-list took an afternoon. Engineers then checked early results against what they meant by distributed systems experience, and the search language was tightened. A check to copy: an engineer reads the first 20 cards and marks which ones show real distributed systems work, then the brief is edited.
3. Reveal verified contacts Reveal verified emails and phone numbers for the engineers worth messaging. A reveal is charged only when a value comes back. Credits: 1 per person a search shows, 4 per email revealed, 12 per phone. Reveal only the engineers you will actually message.
4. Run sequences Email sequences from your own mailbox, from Sourcing Pro up. A reply or opt out stops the sequence, so nobody gets chased after answering. A message to copy: name the system the engineer built, the problem your team is solving, and offer an interview slot they can book tonight.
5. Invite interested engineers to the AI interview Everyone who replied with interest was invited to a live, two-way AI video interview on a technical rubric. Today a role runs at 10 or 20 minutes. Example rubric, ours: data correctness 30, failure handling 30, trade-off reasoning 25, ownership 15. Use 20 minutes so follow-ups have room.
6. Let them book outside work hours Candidates pick any slot inside the window the team opens, without an account. The customer reported most engineers chose between 9 PM and 1 AM. Open evening slots inside the role's window. Employed engineers book after work instead of taking time off for a call.
7. Engineers review reports and run final rounds The team read each report, opened the recording on borderline scores and decided who moved on. Senior engineers met only the finalists. A review rule to copy: replay the recording for any 3 on data correctness or failure handling; a 4 or 5 on both earns a final round.

Why did verified contacts matter for passive backend engineers?

Revealing a verified email or phone for an engineer found in search.

The record credits the 31% interview rate mostly to direct contact details that reach the engineer, rather than guessed work addresses. Contact details are enriched from 30+ sources. A reveal costs 4 credits for an email and 12 for a phone, and is charged only when a value comes back.

What example questions suit an AI interview for a backend engineer?

Examples you could add to a backend role, not the customer's questions. The AI asks added questions in its own order and adds its own. More in our backend developer question bank.

Tell me about a service you owned where two writers could update the same record. How did you keep the data correct?

Names the concrete approach they used, such as versioning or idempotency keys, and explains what failed before they added it.

What happens to your design if one region is cut off for ten minutes?

Walk me through the worst production incident you were on call for in a distributed system.

Gives a clear timeline, separates the trigger from the root cause, and says what they changed afterwards in code or process.

How would you have caught that earlier?

You need a queue between two services. What do you weigh before choosing one?

Talks about ordering, delivery guarantees, throughput and who operates it, then picks one for a stated reason.

Your consumer starts falling behind. What do you look at first?

Other teams will call your service thousands of times a second. What do you decide about the API first?

Starts from the callers: idempotent writes, timeouts, pagination and versioning, and says plainly what the service will not promise.

A caller retries every failed request at once. What happens to your service?

A latency alert fires in the middle of the night and nothing was deployed. Where do you look?

Checks dependencies, traffic shape and saturation before reading code, and names the traces or dashboards they would open first.

It turns out to be one noisy customer. What do you change so it cannot happen again?

What changed for backend hiring, number by number?

420 engineers were contacted, all sourced. 31% of them took an interview, about 130 interviews. From those, 5 were hired in 10 weeks: one hire every 2 weeks, 84 contacted and about 26 interviewed per hire. For scale, Ashby's 2023 data put engineering reply rates at 21%. Here 31% went further than a reply, to a completed interview.

Search to signed offer took 19 days. Against the agency's 6+ week lead time, 42 days or more, that is at least 23 days sooner. It is also under half the 39-day median time to fill in SHRM's 2026 benchmarking (SHRM, read 5 October 2026).

On cost, 5 placements at $25,000 would have been $125,000. Total tooling came to under 4% of that: under $5,000 for 5 hires, or under $1,000 per hire. That leaves more than $24,000 per hire in the engineering budget, against the agency quote.

How was the backend hiring math worked out?

Inputs are customer-reported. Everything after an equals sign is our arithmetic.

INTERVIEWS
420 contacted x 31% = 130.2, about 130 interviews.

FUNNEL PER HIRE
420 / 5 = 84 contacted per hire.
130 / 5 = 26 interviews per hire.
5 / 130 = 3.8% of interviews became hires.
5 / 420 = 1.2% of contacts became hires.

PACE
10 weeks / 5 hires = 1 hire every 2 weeks.

TIME TO SIGNED OFFER
Agency lead time: 6+ weeks = 42+ days.
Actual: 19 days.
42 minus 19 = 23 days sooner, at least.

AGENCY FEES AVOIDED
5 x $25,000 = $125,000.
Tooling under 4%: $125,000 x 4% = $5,000, so under $5,000.
Under $5,000 / 5 hires = under $1,000 per hire.
$25,000 minus $1,000 = more than $24,000 kept per hire.

AGAINST BENCHMARKS
$25,000 / $135,980 BLS median pay = 18.4%.
19 days / 39 days SHRM median = 49%, under half.

What did hiring look like for the engineering lead before and after?

Before follows the record's account of the inbound pipeline. After follows the rollout it describes, with the reported numbers.

All figures customer-reported, except about 130 interviews, which is 420 x 31%.

How did the AI interview test distributed systems depth?

A live AI interview asking a follow-up built from the candidate's last answer, before moving on.

Early in the rollout, some profiles looked strong on backend keywords but were too shallow for the role. The interview is where that shows. The rubric is fixed for the role, the questions adapt live, and a vague answer about consistency or failure gets a follow-up instead of a pass.

What does this backend hiring case study not show?

• One company and 5 hires: a planning data point for specialist backend roles.
• The 6+ weeks and $25,000 are agency quotes, so the comparison is against a search never run.
• The AI interview is a spoken first round; pairing on real code stayed with the team.
• The setup dates from August 2026. The product has changed since, and this page describes today's workflow.

What still needed engineers in this backend hiring setup?

Three things. Engineers had to say what distributed systems experience really meant, because the first searches matched keywords rather than depth. They had to read reports and judge borderline cases. And they ran every final round, where pairing on real code, team fit and the offer conversation happen.

The AI interview is a spoken conversation over video. It is a strong first technical round, but it is not a take-home or a pairing session, and the team still owned that part. The Cognitive is not an ATS, so offers and later interview scheduling stayed in the team's own tools.

Which engineering teams does this approach fit?

Do

• Specialist roles where the right people rarely apply.
• Teams that can write a precise brief and check the first long-list.
• Hiring managers who want senior engineers to meet only finalists.

Do not

• Roles that inbound already fills with strong applicants.
• Teams that want LinkedIn InMail as the main channel.
• Processes that need a coding environment in the first round.

Try it on one backend role

Write the brief in plain English, check the first long-list with an engineer and invite the interested ones to an AI interview. Start free at https://app.thecognitive.io/signup.

Where to go next

Hire a backend developer · AI interviewer for backend developers · Backend developer interview questions · How AI sourcing works · Pricing · All case studies

Sources and notes

Figures reported by the customer; setup dated August 2026. Source: The Cognitive case study record for DevTools SaaS Company, Remote US, created 4 August 2026 and read 4 October 2026. The record does not name the company. Example questions, brief and rubric are ours, and derived figures are our arithmetic, labelled where they appear.

External context, all read 5 October 2026: BLS Occupational Outlook Handbook, software developers (May 2025 pay, 2025 to 2035 openings; page modified 27 August 2026); Ashby candidate sourcing report (2023 data, published 4 March 2024); SHRM recruiting benchmarking (2026 median time to fill).

Frequently asked questions

How do you hire backend engineers who are not applying?

Source them. This DevTools company wrote a plain-English brief, searched ~900M public profiles, revealed verified contacts and ran email sequences. 420 engineers were contacted, 31% took an AI interview and 5 were hired in 10 weeks, none from job boards.

How much does a backend engineer recruiting agency charge?

The agencies this company spoke to quoted $25,000 per placement with a 6-week lead time. Fees vary by agency and salary, so get your own quotes. For 5 hires, that quote came to $125,000. The company reported its tooling cost under 4% of that.

What is a good sourced-to-interview conversion rate?

This team reported 31% of contacted engineers taking an interview, and the record calls that high for cold outreach. It credits verified direct contact details over guessed work addresses. Your rate will depend on the role, the message and how tight the brief is.

Why did engineers interview between 9 PM and 1 AM?

They were employed and not job hunting, so a daytime screening call meant taking time off work. Candidates book any slot inside the window the team opens, and the customer reported most chose late evening. Open evening slots if you hire passive engineers.

Can an AI interview assess distributed systems experience?

It can test how a candidate explains consistency, failure handling and trade-offs, and follow up when an answer stays at keyword level. It is a spoken first round, not a coding exercise. This team kept pairing and final rounds with its own engineers.

How long does it take to hire a backend engineer?

This company reported 19 days from search to signed offer, and 5 hires in 10 weeks overall. SHRM's 2026 benchmarking puts the median time to fill a nonexecutive role at 39 days, so 19 days is under half that. Measure from the day the search starts to the day an offer is signed.

How many candidates do you need to contact per backend hire?

Here it was 84 contacted per hire on our arithmetic: 420 contacted for 5 hires. About 26 interviews led to each hire. A tighter brief usually lowers both numbers, which is why this team tested its first long-list with engineers.

Does The Cognitive send LinkedIn messages?

No. Outreach is email sequences, from Sourcing Pro up, sent from your own mailbox. A reply or opt out stops the sequence.

Does the AI reject engineering candidates automatically?

No. The report offers a suggested verdict, and that is all it is. In this case an engineer read it, replayed the recording when unsure and made the call. Tab switches and similar integrity flags sit in the report for review; they never lower the score.

How much does AI sourcing and interviewing cost?

AI Sourcing and AI Interview each start at $99/month, on monthly plans you can cancel anytime. Sequences need Sourcing Pro or above, or a Team or Business plan. Each AI Interview plan includes a monthly interview allowance, and past it you buy an interview top-up.

How much do backend engineers earn in the US?

The BLS puts median pay for software developers at $135,980 in May 2025 (BLS, read 5 October 2026). Specialists in distributed systems are a slice of that group. For context, this team's $25,000 agency quote equals 18.4% of that median on our arithmetic.

What reply rate should you expect when sourcing engineers?

Ashby's 2023 sourcing data put the average campaign reply rate at 19.6% and engineering at 21% (Ashby, read 5 October 2026). This DevTools team reported 31% of contacted engineers going on to take an AI interview, which it credits to verified direct contact details.

Should the first backend round be a coding test or a conversation?

This team used a spoken first round: a live AI video interview on a technical rubric covering consistency, failure handling and trade-offs, with follow-ups when an answer stayed vague. Pairing on real code came later with its own engineers. The Cognitive has no coding environment, so plan a separate code round if you need one.

Where do you find distributed systems engineers?

Mostly outside job boards, because most are employed and not applying. Write a brief that names the systems they must have run, such as a queue or a distributed data store in production, search public profiles against it, and contact them directly. All 5 of this team's hires came from sourcing.

Explore: Hire software engineers · Hire a Go backend engineer · Passive candidate sourcing · AI recruiting for tech hiring · Technical recruiting tools · Technical interview software