Your first run ranks the whole pile on capability
Paste the role. Set the bar. Press Run. The agent reads every resume for what the person has actually done, not the words they used, and hands back a ranked shortlist with a reason on every line. Start where you are on a Tuesday: 247 resumes, and you, reading them.
Free credits on signup. No demo gate. Scored per resume, multilingual pools.
Read five resumes yourself, and watch the clock climb while 242 wait behind them.
Job description in, must-haves and bar set, Run pressed. Nothing to install.
Every resume read at the same depth, a ranked shortlist, evidence on each line.
Scroll to explore
Act 1 · Run 00 · you do it
Screen them the way you do now
Built the subscription billing service, proration, dunning and refunds, in Go. Handled a Postgres shard split with no write downtime.
Go · Postgres · RabbitMQ · billing · webhooks · gRPC
Five loaded here. The other 242 sit in the folder, each one about this long.
Worth re-reading: Nandita Rao
The job description asks for someone who developed distributed systems. Her resume never uses that word. She writes built, wrote, shipped. Her title reads Software Engineer II, not Senior. On a keyword pass she sinks. On a capability pass she is the strongest person in the pile: a production ledger, a real idempotency layer, 2.1M payouts a day.
Nobody reads resume 200 the way they read resume 2. That is arithmetic, not discipline, and the pile does not care.
Sample data. See it on your own roles.
Open resume. Read. Decide. Next. There are 242 more behind these five.
The strongest candidate in the pile hid behind a verb.
Act 2 · setup · three steps
Three steps, then hand it over
This is the real onboarding, not a mockup of it. Paste a job description, tell it what matters, choose how high the bar sits, filled in with the role you were just screening for.
The JD, the hiring-manager email, the intake notes. The agent pulls the required capabilities straight out of it. Nine get extracted here.
Tap the capabilities that actually matter, then choose how high the bar sits. Matched as capability, so built a ledger counts the same as developed one.
The 247 resumes you were about to read by hand. The agent reads all of them at the depth you gave resume one.
Nothing is sent and nobody is contacted. This runs the sample pile below.
Sample role. Setup is per role and it sticks. Your second role takes about forty seconds.
Role read, must-haves set, bar chosen. One button left.
Act 3 · Run 01 · the agent works
Same pile. Every resume, same depth
- ✓Job description read. 9 capabilities extracted, 3 marked must-have.
- ✓Bar set: evidence of the work in production, with scope.
- ✓Parsing 247 resumes. 31 file layouts, 4 scanned PDFs sent to OCR.
- ✓Normalising verbs. Built, wrote, shipped, developed read as one signal.
- ✓Scoring on capability evidence, not keyword frequency.
- ✓Checking scope on every claim: volume, ownership, on-call, team size.
- ✓2 resumes flagged for a human. Claims outrun the dates on them.
- ✓42 shortlisted, 205 declined, each with a written reason.
- Rohit Anand74%
- Priya Sathe—
- Imran Qureshi88%
- Nandita Rao94%
- Vikas Menon83%
Each score is capability evidence, not keyword frequency. The row being read shows in cyan.
Built a production ledger and the idempotency layer under it. Title says Software Engineer II; the work says senior.
Evidence · “built the ledger reconciliation service that settles 2.1M payouts a day”. JD asks for developed, she writes built. Same capability, different verb.
Owned the payout pipeline on-call and made retries idempotent himself. Platform depth, less service authoring.
Evidence · “cut paging volume 41 percent by making retries idempotent” · “migrated settlement from cron to an event queue”
Real money-adjacent systems, billing, proration, refunds, plus a live shard split. Settlement exposure is thinner.
Evidence · “subscription billing service, proration, dunning, refunds” · “Postgres shard split with no write downtime”
Fraud-scoring pipeline at 9K events a second, event-driven end to end. Ledger work is adjacent rather than owned.
Evidence · “rebuilt the scoring consumer to exactly-once semantics after a duplicate-charge incident”
Keyword-dense, capability-thin. The resume most likely to top a keyword search. Strong API work, no money movement.
Evidence · “developed” appears 6 times; no payment, ledger or settlement path in any role
Sample data. See it on your own roles.
Every resume read for what the person did, not the words they used.
A shortlist you can argue with, because the evidence sits on every line.
What AI Resume Screening does
AI Resume Screening reads every application the way a good recruiter would, weighing skills, context and evidence rather than counting keywords. It scores each candidate against the role on one consistent rubric and ranks the whole pool with the reasoning attached, so resume number 500 gets the same scrutiny as resume number 1.
You paste a role, set the must-have capabilities and the bar, and the agent reads the pile in minutes. Shortlisted candidates move straight into the next step, an AI phone screen, an assessment or a structured interview, and every decline keeps a written reason, so nothing disappears silently.
Capability, not keywords
Built a payments service in Go counts toward backend experience even when the phrase backend engineer never appears.
One rubric, ranked pool
Every candidate is scored against the same role requirements, producing a ranked shortlist with a per-candidate match reason.
Evidence on every line
Each score cites the line in the resume it came from, so your team and your auditors can inspect any rank.
Metered per resume
Screening is scored per resume, so a pile of hundreds is read at the same depth as a handful.
Multilingual pools
Differently-languaged resumes compete fairly in one ranked pool.
Connected to the funnel
Results land in the Inbuilt ATS and chain to the Calling Agent, AI Assessments and the AI Interviewer under one candidate record.
What teams say
Sample quotes shown for layout review. Real customer stories will replace these.
It reads for capability, not keywords. Candidates we would have filtered out on a job title got surfaced because the agent saw what they had actually built.
Two hundred and forty seven resumes used to be a morning I dreaded. Now it is a three minute run and a shortlist I trust.
Every score comes with the line it came from, so my hiring managers stopped second guessing the shortlist.
Frequently asked questions
How does AI resume screening rank candidates?
One consistent rubric is applied to every applicant, scoring capability in context rather than matching keywords, and each score comes with the reasoning behind it. You get a ranked shortlist you can inspect line by line, and your team always makes the final advance or reject decision.
Does it work for non-technical roles?
Yes. The agent scores against whatever competencies the role defines, whether that is sales, support, operations, engineering, or fresher hiring where potential matters more than history.
Can it screen resumes in multiple languages?
Yes. Multilingual understanding lets differently-languaged resumes compete in one ranked pool, so you can evaluate talent fairly across regions.
Does AI resume screening integrate with our ATS?
Yes. Applications can flow in from a connected ATS or HRMS, results land in the Inbuilt Agentic ATS or sync back through integrations and webhooks, and screening decisions stay inside the system your team already uses.
What happens to candidates who are not shortlisted?
They stay in your talent pool. The AI Sourcing Agent keeps every parsed resume searchable, so today’s runner-up is discoverable the moment a better-matched role opens.
How quickly are screening results available?
Ranked results arrive in minutes. Hundreds of resumes are processed concurrently, each with its score and reasoning, so recruiters open a sorted shortlist instead of a raw pile. Sign up, get free credits on us, and screen a live role before you commit to anything.
Works with the rest of the platform
Bring a pile. Run it once.
Set AI Resume Screening up on your live job description, run it against a real stack of resumes, and keep the ranked shortlist with its reasons whatever you do next. Humans keep the final call.