Paste a job description and TalentGraph designs a fair, job-relevant interview, runs it in the candidate's browser with dynamic follow-up questions, and returns a scored screening report with transcript-level evidence. The AI never makes the hiring decision — you do.
Sent as X-API-Key with recruiter requests. Stored only in this browser.
Create a job and invite a candidate to see sessions here.
The first round is where most hiring inconsistency is introduced.
Whoever is free that afternoon no longer determines what a candidate is asked. Two people for one role get a conversation you can actually compare.
The screen is generated from the structured job, so it probes the skills the role actually needs instead of a generic template.
Every answer gets a score and the notes behind it, so a hiring manager reading the summary can see why — not just what.
The output is a record, not a recollection.
{
"title": "Backend Engineer II",
"required_skills": ["Python", "FastAPI",
"SQL", "Docker"],
"min_experience_years": 3,
"seniority": "mid"
}SCREEN — Backend Engineer II
5 questions · ~20 minutes
Q2 "Walk me through how you'd keep a retried
payment from charging twice."
probes SQL · idempotency · seniority
ANSWER SCORE 4 / 5
· Named idempotency keys unprompted
· Described the DB constraint correctly
· Did not cover the failure window
SUMMARY
Python 4/5 · FastAPI 4/5 · SQL 4/5
Docker 3/5 · communication 5/5
Recommend: advance to technical roundGenerate the screen from the role, run it, keep the record.
1 Pick the structured role to screen for 2 Screen generated .............. 5 role-specific questions 3 Candidate answers, question by question 4 Each answer scored with the reasoning recorded 5 Summary produced ............. per-skill scores + recommendation 6 Compare candidates side by side on the same scale
The free tier needs no card. If you would rather see it on your real requisition load first, we will walk through it with you.