Platform

Everything between a job description and a hiring decision

Every part feeds the same scorecard. Use all of it, or start with the piece you need most.

01The Test Agent · AI Test Generation

Turn a job description into a test

Hand the Test Agent a job description. It finds the skills the role really needs, proposes the shape of the test, and writes the questions — and nothing reaches a candidate until you have approved it.

A PDF job description, or a few lines of requirements
You confirm the skills, then the shape of the test
Regenerate any single question with a nudge
Generated questions are labelled, so reviewers know what to check

Assessment blueprint

AI proposed

✓Describe role
✓Review skills
3Blueprint
4Generate

ROLE

Design Verification Engineer

from DV_Engineer_JD.pdf · 8 skills found

QUESTIONS

MINUTES

TOTAL MARKS

CUT-OFF

SKILLS

Difficulty

Easy25

Medium50

Hard25

Question type mix

36 questions

MCQsingle answer2040 marks
MCAmultiple answer824 marks
SubjectiveAI-evaluated648 marks
CodingPython scripting230 marks

Skill weightage

re-proposing…

SKILLWEIGHTQS

SystemVerilog22%8
UVM methodology20%7
Coverage & assertions (SVA)14%5
Digital design fundamentals12%4
Constrained-random verification10%4
Protocols (AXI, APB, PCIe)10%4
Debug & waveform analysis7%2
Scripting (Python, Tcl)5%2

Steer the AI: tell it what to change

Sample data

How it’s set up

1
Upload the job description
A PDF works, and so does plain text typed into the box.
2
Confirm the skills, then the blueprint
You confirm the skills, then the shape of the test. Nothing is written until you agree.
3
Review the drafted questions
Regenerate any single question with a hint. Generated questions are labelled in the test, so reviewers know which ones to check.

How it’s set up

1
Upload the job description
A PDF works, and so does plain text typed into the box.
2
Confirm the skills, then the blueprint
You confirm the skills, then the shape of the test. Nothing is written until you agree.
3
Review the drafted questions
Regenerate any single question with a hint. Generated questions are labelled in the test, so reviewers know which ones to check.

02The Assessment Agent · Skill Assessments

Real code, real queries, real diagrams

Build one assessment from every question type the role needs. Candidates write and run real code, query a live database, draw system diagrams, and explain their reasoning in their own words.

Coding in 10 languages, run against real test cases
SQL against a real MySQL database
Diagrams, written answers, MCQ and more
AI evaluation you can override
A shared question library, tagged and filtered

Front-end Engineer · question 4 of 18

Built from the job description

34:20Submit test

Coding · 10 languages15 marks

Build a searchable product catalogue across three files. It must filter as the user types.

Explorer

src

components

ProductCard.tsx

SearchBar.tsx

hooks

useDebounce.ts

Catalog.tsx

types.ts

Catalog.tsxSearchBar.tsx▶ Run tests

1import { useMemo, useState } from "react";

2import { SearchBar } from "./components/SearchBar";

3import { useDebounce } from "./hooks/useDebounce";

5export default function Catalog() {

6 const [query, setQuery] = useState("");

7 const [category, setCategory] = useState("All");

8 const search = useDebounce(query, 250);

10 // filter once per change, not on every render

11 const visible = useMemo(() => {

12 const term = search.trim().toLowerCase();

13 return PRODUCTS.filter((p) =>

UnsavedReact 18Ln 1src/Catalog.tsx

Live preview running

Products

Search productsAllAudioWearables

6 results

Aria Wireless HeadphonesAudio$129Pulse Smart WatchWearables$199

Tests0 / 6Console0 / 15 marksnot run yet

renders every product as a card8ms··/3

filters by name, case-insensitively12ms··/3

debounces the search by 250 ms31ms··/2

category chips filter the grid9ms··/3

result count matches the cards6ms··/2

empty state when nothing matches5ms··/2

Similarity 4%no match in your question librarytyped over 14 min · no paste events

Sample data

How it’s set up

1
Pick or write the questions
Pull questions from your own bank, or write new ones in whichever format the role calls for.
2
Set the rules
Set the length, the weighting and how closely the session is watched.
3
Invite candidates
Invite one candidate or a thousand. Each gets their own link and takes the test in the browser.

How it’s set up

1
Pick or write the questions
Pull questions from your own bank, or write new ones in whichever format the role calls for.
2
Set the rules
Set the length, the weighting and how closely the session is watched.
3
Invite candidates
Invite one candidate or a thousand. Each gets their own link and takes the test in the browser.

03Spoken answers · SpeakPro

Hear how they communicate before the interview

SpeakPro records spoken answers and scores them on the dimensions that matter for client-facing and support roles. Recruiters get the recording, the transcript and a score they can adjust.

Read aloud, repeat, listen, or answer freely
Scenarios for support, sales and more
Scored by dimension, with the recording
Mic and speaker checked before the test

Guest recovery: the lost booking

SpeakProOpen response✦ AI-evaluated

0.0 / 10 marks

PROMPTA guest arrives at 11 pm after a long flight. Their booking can’t be found and the hotel is nearly full. Respond as you would at the front desk.

Candidate video1.25×

“I can see you’ve had a long journey, and I’m really sorry.”

0:00 / 1:24

0.0

STRONG

Calm, empathetic recovery with a concrete guarantee. A few fillers while she searches for the next step.

PACE

0wpm

Natural

FILLERS

0.0%

“um”, “so”

PAUSES

None over 2s

CLARITY

Very clear

Dimension analysis

overall 4.3 / 5

Task fulfilment4/5

Empathy5/5

Professional tone5/5

Grammar & vocabulary4/5

Fluency3/5

Intelligibility5/5

Transcript & evidence

stood out watch for

✦ TalentDecrypt Communication AI · the recruiter can adjust the score

Sample data

How it’s set up

1
Choose a task style
Open response, read aloud, repeat a sentence, or listen and then answer. Pick a scenario such as customer support, sales or technical explanation.
2
The candidate checks their audio
A microphone and speaker check runs before the test, so a bad headset shows up before the recording does.
3
Review the evaluation
Hear the answer, see the moment behind every score, and change the score if you disagree.

How it’s set up

1
Choose a task style
Open response, read aloud, repeat a sentence, or listen and then answer. Pick a scenario such as customer support, sales or technical explanation.
2
The candidate checks their audio
A microphone and speaker check runs before the test, so a bad headset shows up before the recording does.
3
Review the evaluation
Hear the answer, see the moment behind every score, and change the score if you disagree.

04The AI Interviewer · AI Interviewer

A first-round interview for every candidate

The AI Interviewer runs a structured voice interview from a question plan you approve. It asks follow-up questions based on what the candidate said, and produces a transcript, a recording and scored competencies.

A question plan you approve first
Follow-ups that react to each answer
Candidates join from a link, at any hour
Recording, transcript and scored competencies

Backend Engineer · first-round interview

1Plan
2Interview
3Verdict
4Evidence

Recording

Who are you hiring?

30 minutesVoiceFour topicsAdapts to answers

The plan is yours before anyone is invited. Change a topic, push one deeper, or drop it.

Approve plan

Interview plan drafted from the role

Background & ownershipbrief
System designgo deep
Debugging under pressurego deep
Working with othersbrief

Weighted to the role · edit, reorder or drop any topic

Arun ShettyBackend Engineer · AI interview · 24 minCompleted

Recommended: proceed to a live round

Strong system-design depth and hands-on debugging; communication is clear. Probe stakeholder trade-offs live before a final decision.

Strong: system designStrong: debuggingProbe: stakeholder trade-offs

Interview recording24:00

04:12 Strongest answer17:40 Worth probing

✓ AdvanceHold for reviewDo not proceed

A recommendation — you decide. The AI never rejects a candidate.

Strength profile

five dimensions, against the role benchmark

Candidate Role benchmark below it

Ranks inTop 22%of 46Topics3 of 488% coveredHe talked62%the AI asked

Evidence by competencybar = candidate · ◆ = role average

System design & scalabilityFull coverage8.6

04:12“I’d switch the nightly batch to an event-driven matcher keyed on settlement id, so a replay can’t double-count.”

Debugging & production ownershipFull coverage8.0

11:20“We double-charged about 200 customers. I rolled it back in ten minutes, then reconciled the duplicates before support got the first call.”

Stakeholder & cross-team trade-offsPartial5.8

17:40“You’d sit down with them and find the middle ground, I suppose.”

Every score links to the moment it came from · click a quote to hear it

Strengths

Designs for correctness under load — idempotency, watermarks, back-pressure

Owns production issues end to end

Probe further

Stakeholder trade-offs stayed abstract — no concrete example of navigating conflict.

Take into the human round

Tell me about a time the payments team wanted something that conflicted with your plan. What did you give up?

Walk me through a decision where you were overruled — how did you handle it?

Copy to notesRecording · transcript · integrity attached

Sample data

How it’s set up

1
Generate a question plan
Describe the role and the platform drafts the interview questions, grouped by the competencies you want to hear about.
2
Edit the plan
Reword, reorder, add or remove questions. The interviewer asks what you approved, plus follow-ups on what the candidate says.
3
Send the link
Candidates join from a link at a time that suits them. The recording, transcript and competency scores land in their report.

How it’s set up

1
Generate a question plan
Describe the role and the platform drafts the interview questions, grouped by the competencies you want to hear about.
2
Edit the plan
Reword, reorder, add or remove questions. The interviewer asks what you approved, plus follow-ups on what the candidate says.
3
Send the link
Candidates join from a link at a time that suits them. The recording, transcript and competency scores land in their report.

05CheqMate AI · CheqMate AI

See what ordinary proctoring misses

CheqMate AI watches beyond the browser. Screen, webcam, device-level signals and TriNetra’s second camera come together into one evidence-backed view of every assessment — and every flag keeps the recording behind it.

TriNetra: the candidate’s phone as a second camera on the desk and the room
A locked test window, watched for what else is running
Screen, window-focus and face-detection events, tied to the question
A live wall with chat, and ID checked on a call before the test starts
Every flag keeps its recording, snapshot or log

talentdecrypt.com/proctoring

Rohan Mehta

Backend Engineer test · question 12 of 18 · 00:49:51

ID verified by greeterCheqMate AI on

Face detected

Webcam

Screen

Someone entered the room and stood beside the desk. Only the phone camera could see them.

Sample data

How it’s set up

1
Choose the level for this test
Webcam and screen for a routine test. Add TriNetra’s second camera, the locked desktop window or a live ID check when the stakes are higher.
2
The candidate links a phone
During the system check the candidate scans a QR code, and the phone becomes the second camera. A greeter can check ID and the room before the test starts.
3
Watch live, or review afterwards
Your team sees every feed on one wall and can message a candidate. Each flag is saved against the question they were on.

How it’s set up

1
Choose the level for this test
Webcam and screen for a routine test. Add TriNetra’s second camera, the locked desktop window or a live ID check when the stakes are higher.
2
The candidate links a phone
During the system check the candidate scans a QR code, and the phone becomes the second camera. A greeter can check ID and the room before the test starts.
3
Watch live, or review afterwards
Your team sees every feed on one wall and can message a candidate. Each flag is saved against the question they were on.

06The Scorecard Agent · Candidate Scorecards

One scorecard, with the evidence behind every number

Each candidate gets one scorecard: scores by skill, the integrity record, spoken-answer evaluation and the interview. An AI summary explains the verdict in plain language, and every score links to the answer it came from.

Every score links to the answer behind it
An integrity record with snapshots and flags
An AI summary written from real results
A communication profile for spoken answers

Tests / Backend Engineer, hiring assessment / Sneha Menon

Sneha Menon

Backend Engineer

Finished

OVERALL SCORE

0 / 100

Strong result

PERCENTILE

0th

Rank 13 of 214 candidates

PROCTORING COVERAGE

Webcam100%

Screen100%

Phone97%

SESSION

Invited: 12 Sep, 10:04 am
Started: 14 Sep, 7:30 pm
Finished: 14 Sep, 8:34 pm

Decision pending.

AI insight · TalentDecrypt AI RecommendedTop 6% of 214 candidates · clean integrity record

Strong in API design & SQLBoth coding problems solvedFinished 11 min earlyNo integrity flags

OVERALL

0/100

+26 vs cohort

ACCURACY

11 of 12 correct

TIME USED

0/75 min

Finished 11 min early

INTEGRITY

Clean

0 flags, 3 feeds

Skill analysis

Sneha Cohort average

Sample data

How it’s set up

1
Results arrive on their own
Scores, recordings, flagged events and the interview are added to the scorecard as each part finishes.
2
Open any number
A skill score opens the answers behind it. The integrity tab lists each flagged event with the question it happened on, beside webcam and screen snapshots.
3
Adjust, then decide
Override a score where you disagree with the evaluation. The AI summary is written from the candidate’s actual results.

How it’s set up

1
Results arrive on their own
Scores, recordings, flagged events and the interview are added to the scorecard as each part finishes.
2
Open any number
A skill score opens the answers behind it. The integrity tab lists each flagged event with the question it happened on, beside webcam and screen snapshots.
3
Adjust, then decide
Override a score where you disagree with the evaluation. The AI summary is written from the candidate’s actual results.

See it with one of your own roles

Tell us a role you are hiring for. We will build the test, run a candidate through it and walk you through the scorecard that comes out.

Book a demo info@talentdecrypt.com