For Indian IT and tech recruiting teams
Find the three real engineers hiding in200 look-alike resumes
It’s 6pm, the backend req has 214 applies, and you’ve opened 40 of them. Recruitkar scores every one against your actual JD and leaves a ranked shortlist waiting when you sit down tomorrow.
What a tech req actually costs you
The same week, twice
One backend req, two versions of the same week
You download resumes one at a time, rename the files, paste names into a sheet. Two hours gone before you have read a single line.
Tick the applicants on the Naukri, Indeed or Shine tab and the browser extension pulls them in at 0.5 credit each, already de-duplicated against everyone on the job.
You Ctrl+F for Kubernetes, Kafka, React. Everyone has the words, because everyone read the JD before writing the CV. So you go by company tier and gut.
Every profile carries a score against your JD, a verdict from strong fit to not fit, matched and missing skills by name, red flags, and questions to ask that person.
Forty numbers copied into WhatsApp one by one, the same message typed forty times, and no record of who replied on which day.
Bulk email with merge fields, WhatsApp templates, or an AI voice call, run off the same ranked list. Email is 1 credit per candidate ever, not per message.
Thursday and Friday on the phone asking the same six questions: current CTC, expected CTC, notice period, offers in hand, location, comfortable with the stack.
An AI voice or video interview in the candidate’s own language, recorded, with a scorecard and a hiring recommendation sitting on the card when you open it.
Five profiles pasted into an email at 11pm, and a week of chasing the hiring manager for a yes or a no.
A teammate reviews internally first. Then it goes out with only the fields you chose to show, and the client’s approve, reject or note comes straight back onto the candidate.
The part only tech recruiters deal with
Judged on the real stack, not the keyword
“5 years React” is not information. It could be five years building a design system four teams depend on, or five years editing JSX somebody else architected. The resume reads identical either way. Keyword matching makes it worse, not better: it rewards whoever read the JD before writing their CV, and it buries the engineer who wrote “frontend” where you wrote “ReactJS”. Recruitkar reads the profile against the job, not against a word list.
Skills is one dimension out of about ten
Experience, seniority, title, industry, nice-to-haves and overall relevance are each scored on their own. So listing “microservices, system design, mentoring” does not carry anyone past the seniority read. The stack words and the depth behind them are two different questions, and they get asked separately.
You get the reasoning, not a number
Matched skills and missing skills side by side, the red flags worth raising, and suggested interview questions for this specific person. Your tech panel walks in with something to probe instead of a CV and a shrug. And when you disagree with the verdict, you can see exactly what it read.
A thin JD gets no score at all
Paste a two-line req, or import a CSV row that is just a name and an email, and Recruitkar will not hand you a confident 82 percent. Those dimensions come back empty instead. A score you cannot trust is worse than no score, so it refuses to make one up. The flip side is honest: put in a vague JD, get vague output.
In practice
How it works, on a normal Tuesday
Set the req up properly, once
Add the client, the department, the job. Each client sits in its own space, so nothing bleeds between accounts. Then paste the real JD, not the two-liner the hiring manager sent on WhatsApp, because everything after this is scored against it.
Pull candidates in from wherever they already are
Import your Naukri, Indeed or Shine applicants with the browser extension. Upload the resumes sitting in a folder. Drop a CSV. Pull a LinkedIn profile. Share an apply link. Or let AI discovery search external profile databases for you. It all lands in one ranked pipeline with duplicates merged into one card.
Work the top of the list instead of the whole list
Read the verdicts and the red flags, then email, WhatsApp or voice-call from the same screen. Move people through the seven board stages, from new to shortlisted to hired. Get a senior on your team to check a profile before it reaches the client.
What it costs, in credits
Straight answers
Will this work for me?
“All my candidates are on Naukri. I am not going to start sourcing from some database I have never heard of.”
Then don’t. Install the browser extension and import the applicants already sitting on your Naukri, Indeed or Shine req at 0.5 credit each. They get scored exactly like everything else. AI discovery is available if you want it, not a step you have to pass through.
“I have seen AI matching before. It gives everyone 85 percent and I still end up reading all of them.”
That is what the quality gate exists to stop. If the JD is two lines, or the CSV row is a name and an email, those dimensions come back null rather than a flattering number. You will see fewer scores, not more. The honest limit: the output is only as good as the JD you paste. A vague req produces vague verdicts, and it will show you that instead of hiding it.
“If a model decides who is a good fit, I am handing my judgement to a black box and I will get blamed for it.”
You get the reasons, not just the number: matched skills, missing skills, red flags, and the questions to ask. Nobody is auto-rejected and nothing is sent without you. It orders the list. You decide who gets the call, and you can see what it read to arrive at the order.
“My team will fire off half-finished profiles to a client and it will land on me.”
Switch on internal review. A member’s candidate cannot be shared with a client until a teammate approves it, in whichever direction your team works. You can also set a per-member credit cap so nobody spends the month’s budget by Wednesday.
“An AI interview will insult a senior engineer and he will drop out of the process.”
Reasonable worry, and the answer is to use it as a screen, not as the panel. It runs in 11 or more languages, records, logs proctoring events, and returns a scorecard with a recommendation. When a profile deserves a human, schedule one: Google Calendar or Outlook, on Meet, Teams, Zoom, or in-office.
Questions
Asked before you ask them
I sourced this developer for a Bengaluru req last month and a Pune req just opened. Do I pay again?
No. Reusing a candidate on another of your own jobs is free. They get re-scored against the new JD, so the same person can be a strong fit on one req and borderline on the next. Same profile, different verdict, no second charge.
The same person applied on Naukri, is in my resume folder, and came through the apply link. Will I see him three times?
Once. The pipeline de-duplicates across every source, so those three arrivals merge into one card with one score and one history. You will not call him twice and he will not get the same blast twice.
We run six clients at a time. How do you keep them from mixing?
The structure is client, then department, then job. Each client keeps its own jobs, candidates and board. A candidate you worked for one client does not surface inside another client’s pipeline unless you put them there.
If I email 300 candidates at once, what actually happens?
It goes out as a bulk send with merge fields, across a deliverability pool, with bounces monitored so you can see what failed. Two limits worth saying plainly: it is a blast tool, not per-candidate AI-written personalisation, and there is no domain warming service bundled with it.