Upload your candidates. Let AI match, interview, and grade them.
Upload a stack of resumes. Screen them against the job. Run proctored interviews with a recruiter in the loop, then share one report. Every score is backed by the transcript and the code.
Set up the job, then screen the stack.
You pick the skills and set a weightage. You set up the JD. Every resume is scored against both, so your interview time goes to the people worth talking to.
Set up the job, then upload
Job details takes the title and the location. Add the JD, pick the skills, set a weightage on each. Then upload the resumes in bulk.
Screen against the role
A resume screening round scores every resume against your JD and your weighted skills. The reasoning sits next to the score, so you see why someone ranked where they did.
Shortlist the strongest fits
Sort by score, read the reasoning, shortlist. You start interviews with a shortlist instead of an inbox.
Send them into the next round
Add the next round and set a promotion threshold. Auto-promote and auto-reject are settings you turn on yourself.
Every step, with screenshots. The guides walk through the same flow, field by field.
Five modes. Eight interview types.
A round runs in one of five modes. If you pick AI interview, you then pick the type. Every round on a job scores against the same skills, the same weightage and the same JD, so the results stack.
The five modes
- Resume screeningNo interview. Reads resumes against the JD and your weighted skills.
- AI interviewEthan runs it. Pick one of the eight types below.
- Human-led interviewYou run it, here or on Zoom, Meet or Teams, with the notetaker in the room.
- AI prompt assessmentCandidates write prompts for real tasks, scored on eight dimensions.
- Coding testSandboxed, real execution against hidden test cases.
The eight AI interview types
| Type | Time | What it covers |
|---|---|---|
| Coding | 20 + 10 min | Programming questions with a code editor |
| General Interview | 15 + 10 min | Broad topics and overall fit |
| Communication | 15 + 10 min | Clarity and articulation |
| Problem Solving or Aptitude | 15 + 10 min | Logical and analytical reasoning |
| System Design | 15 + 10 min | Architecture, with whiteboard support |
| Academic | 15 + 10 min | Faculty recruitment at universities |
| HR Round | 10 + 10 min | Behavioural questions and cultural fit |
| Task | 10 + 10 min | One focused coding task plus 2 to 3 follow-ups |
Times read as round plus intro, so 15 + 10 min is a 15 minute round with a 10 minute intro. Difficulty runs from very easy to very hard, and you set it per round.
Scored on what they said, wrote, and built, not on a face scan.
Every score traces to a transcript, live code, and a structured rubric. Integrity signals go to a human reviewer. Never an automated pass/fail. A human can silent-listen or take over in real time. Built for auditability, with jurisdiction-specific compliance guides for NYC LL144, Illinois, California, and the EU AI Act.
Every score traces to
Transcript
Every score links back to what the candidate actually said, line by line.
Live code
What they wrote and ran in the editor, captured as it happened.
Structured rubric
Consistent, rubric-based scoring against the criteria set for the role.
A human stays in the loop
Integrity signals, reviewed by a human
Integrity signals from the round go to a human reviewer next to the transcript and the code. Never an automated pass or fail.
Silent-listen or take over in real time
A recruiter can observe a live interview without interrupting, then step in and take over whenever they choose.
Server-side resume masking
When reports are shared, identifying details on the resume are masked server-side before the report leaves the pipeline.
Score the calls you are already running.
Not every interview is AI-conducted. Scheduled calls, ad-hoc rooms and old recordings all end up in the same report format.
Calendar notetaker
Send notetaker puts a bot in a scheduled call, or set auto-record once and choose all meetings, internal only, internal excluding 1:1s, or manual. It records, transcribes and scores against the job.
Participants are notified that the call is being recorded.
Instant Meetings
Start instant meeting opens a room when a conversation happens without a calendar invite.
Recorded, then returned as a titled report.
Score a past call
Already ran the call? Score that recording against a job and get the full report without redoing the interview.
The score is backed by the transcript.
Capture the profile. Then hide who it belongs to.
Open a LinkedIn, Naukri or Monster profile and pull it into your talent pool from the panel on the page, with the source recorded against it.
It also masks the name, the contact details and the photo on a resume, so reviewers read the work before anyone enters a pipeline. Phone numbers, emails and links are found by pattern and by context, so they come off two column and graphic heavy layouts too.
- Name
- Candidate A
- •••••••@•••••
- Phone
- ••• ••• ••••
- Photo
- removed
Skills, experience and projects stay intact. Only the identifying details go.
One report, shared on your terms.
The report maps the score back to your skills and your JD, and carries the transcript and the code behind it. It is never an automated pass or fail.
Strong Flask expertise, one unfinished task.
Nine years on Python backends, mostly Flask and REST. Built the search route cleanly, then stopped short of the Elasticsearch query and said so.
- Flask structure modular routes, CORS, jsonify
- Production caching Redis layer, eviction reasoning
- Elasticsearch query and relevance sort unfinished
- Distributed locks named deadlocks, not the fix
What you are looking at
One candidate report. A score per skill, the written summary with strengths and areas for growth, the recording and transcript, the detected events, and a resume score on its own axis. Overview, Analysis, Submissions and Resume are the four tabs. Coded product UI, sample data.
Share one report
A hiring manager or client gets one report with the scores, the transcript and the code from the interview.
Choose who can view
You choose who can open a report: anyone with the link, your organisation, or invited emails only. Reviewers open a controlled link instead of a file forwarded around.
Masked report PDF
Export a PDF with the identifying details removed, so you can share it wider without naming the candidate.
Masking happens server-side. Identifying details are removed before the report leaves the pipeline. Your candidates stay yours, not listed in a shared marketplace.
Chat with Mia. In the app, or from Claude.
Mia sits in the workspace and runs the desk in plain language: put a candidate on a role, chase a pipeline, pull a report. A secure, OAuth-protected remote MCP server exposes the same tools to your own AI assistant. ChatGPT support is in beta.
Financial, team-admin, and destructive actions are intentionally excluded. They stay where a human confirms.
- Type
- Remote MCP server
- Auth
- OAuth
- Exposes
- Dozens of recruiting tools
Excluded by design
- Financial actions
- Team-admin changes
- Destructive operations
Upload your first stack of candidates today.
Screen, interview and grade in one place. A human stays in the loop, and every score is backed by the transcript and the code.
Submit stronger candidates, faster.
Hold every client to the same bar, and send each candidate with the evidence attached.