Superset Assessments

An assessment suite for all your hiring needs

Cognitive, technical, coding, communication and behavioural assessments — built and calibrated by Superset, proctored end to end, and scored straight into your shortlist.

Get started See the formats

Three Superset assessment surfaces: a graduate screening paper being assembled in the Assessment Builder with a question open for editing, a drive's results filling in stage by stage, and the proctoring console with signal counts and candidate camera tiles.

Graduate screening · v3 Freshers · 90 min · 129 marks Draft

Sections

1 Quantitative aptitude 12 questions · 24 marks
2 Logical reasoning 10 questions · 20 marks
3 Core CS 15 questions · 45 marks
4 Coding 2 problems · 40 marks
+ Add section

Question bank Yours and Superset’s

Question 9 of 15 Medium

Which index makes range queries on a sorted column fastest?

A B-tree index on the column
A hash index on the column
A bitmap index on the column
No index; a full scan is cheaper

+3 correct −1 wrong Databases

Graduate drive 312 in session · started 10:00 Live

312 Verified 14 Tab switches 3 Second person

Ananya Sharma

Karthik Rao Tab switch × 3

Arjun Nair

Fatima Khan

Rahul Verma Second person

Meera Pillai

Priya Menon

Rohit Das

Graduate drive results 12,480 invited Results live

Cognitive 12,480
Technical 11,904
Behavioural 10,610
Coding round 9,984

Assessments a year: 12 MillionAssessments a year
Concurrent test takers, tested: 1 LakhConcurrent test takers, tested
Assessment formats: 24+Assessment formats
Programming languages: 65+Programming languages

The catalogue

An assessment tool for every role

Superset supports more than 24 types of assessments. Pick any one, or a combination, for the role — a cognitive screen for volume, a domain test for skill evaluation, or a fully integrated work environment to test the person on the real job.

Verbal, logical and numerical reasoning, with listening and spoken-English sections where the role needs them. The screening layer for volume hiring, scored the moment a candidate finishes.
From IT and data to core engineering (22+ branches), BFSI, semiconductors, pharma and telecom — each calibrated to fresher-level expectations rather than borrowed from lateral screens. Language proficiency alone covers 42+ programming languages.
In-browser coding auto-graded against test cases with time and memory limits, plus debugging and refactoring exercises, front-end simulators and take-home project reviews. The same editor powers paired live-coding interviews with integrated video.
Spoken, written and listening English evaluated by AI — fluency, pronunciation, grammar, vocabulary, filler words and pacing — scored consistently across every candidate before an interviewer ever hears them.
Next-generation work-sample tests that evaluate what a person does, not what a resume claims. Candidates run a business day — seeking evidence, making calls under a budget — and are scored on decision quality, prioritisation and reasoning.
Personality and behavioural batteries, situational-judgement tests and role-specific profiles (sales roles included), reported as workplace signals a recruiter can act on rather than raw trait jargon.
Formats built for engagement as much as evaluation: hackathons, live quiz events, multi-question case studies, game-based assessments and portfolio reviews — run as standalone events or as rounds inside a drive.

One candidate's cognitive and aptitude report: verbal, logical, numerical and listening sections with a question-by-question right, wrong and skipped strip, the total against the team's cut-off, and time spent per section.

Arjun Nair Cognitive and aptitude · 52 min Shortlisted

Right Wrong Skipped

Verbal reasoning 13 / 15
Logical reasoning 11 / 15
Numerical reasoning 8 / 15
Listening comprehension 12 / 15

Total 44 / 60 Cut-off 36, set by your team

Time per section Verbal10:40 Logical13:55 Numerical15:00 Listening8:10

The technical domain test catalogue: search, industry filters from IT and software to semiconductors, the IT tests listed, and the SQL test's sections with an Add to drive action.

Domain tests 94+ tests across 16 industries

Search tests, skills or roles

IT and softwareData and AICore engineeringBFSISemiconductorsPharma +10

Java programming 40 min · MCQ and coding
SQL and databases 30 min · MCQ and query writing
Data structures and algorithms 45 min · MCQ and coding
Debugging in Python 30 min · fix the failing code
Front-end with React 40 min · in-browser simulator
Cloud fundamentals 25 min · MCQ

Fresher-calibrated SQL and databases IT and software 30 min 25 questions 4 sections

1Querying basics10 questions
2Joins and aggregation8 questions
3Schema design5 questions
4Write the query2 tasks

Preview Add to drive

A coding round in the in-browser IDE: the candidate typing a solution, the problem with its time and memory limits, and the test cases passing as it is auto-graded.

two_sum.py tests.py Python 3 · auto-graded

1def two_sum(nums, target):
2    seen = {}
3    for i, n in enumerate(nums):
4        rest = target - n
5        if rest in seen:
6            return [seen[rest], i]
7        seen[n] = i
8    return []
9
10
11# O(n) time, O(n) space

Run 8 / 8 tests passed in 0.02 s

Two sumEasy

Return the indices of the two numbers in the list that add up to the target.

1 s time limit 256 MB memory

Test cases 8 / 8 passed

Example 1 0.01 s
Example 2 0.01 s
Duplicates 0.01 s
5 hidden cases 0.02 s

An Eloquence AI English report for one candidate: a spoken response's waveform with filler words marked, pace and pauses, fluency, pronunciation, grammar and vocabulary scores, an overall score, and the transcript with grammar notes.

Priya Menon Spoken English · 3 responses · 6 min

Response 2 · Describe a project you are proud of

1:24

132 words a minute 2 filler words 1 long pause

Fluency82

Pronunciation76

Grammar68

Vocabulary80

Written 74 Listening 81

77of 100 Overall

Transcript

So the project was, um, the portal for our campus fest. I leadled a team of four, and we shipshipped it in six weeks, like, before the fest opened.

2 grammar notes 2 filler words

A Skill Studio work simulation as CEO of a quick-service food brand: the brief, six months of slipping same-store sales, deliverables completing, the candidate's locked decision, and their decision-quality, prioritisation and reasoning scores.

CEO for a quarter Skill Studio · quick-service food brand 236 in progress

Same-store sales are slipping and a discounting war is on. Decide what to fix first, with the evidence to defend it.

Same-store sales, indexed to April

100 Apr 98 May 95 Jun 93 Jul 89 Aug 86 Sep

Evidence review 12 sources · reviewed
Pricing decision Locked in
Turnaround memo Written response · drafting

Decision locked

What do you fix first?

Delivery times, before discount depth

Late orders are driving repeat customers away; matching the discount war only deepens the margin loss.

Scores so far

74 Decision quality 81 Prioritisation

A behavioural profile from a sales-roles battery: six traits on a radar, each reported as a workplace signal, what to probe in the interview, and the situational-judgement result.

Rohit Das Sales roles battery · 25 min Report ready

Achievement drive82 Sets stretch targets and keeps score
Persuasion76 Changes minds with evidence
Resilience71 Recovers quickly from a lost deal
Customer focus69 Asks before pitching
Team orientation54 Prefers to close alone
Self-discipline62 Needs structure for admin work

Probe in the interview

“Tell me about a deal you won only because someone else on the team helped.”
“Walk me through how you keep your pipeline notes up to date.”

Situational judgement 7 of 8 scenarios answered with the strongest response

Engagement-led formats together: a hackathon mid-event with its quiz, case-study and build rounds and a leaderboard of campus teams, a live quiz with the room's answers, a game-based test, and a portfolio scored against a rubric.

Helio CodeSprint Hackathon · 1,842 teams

✓ Online quiz Round 1
✓ Case study Round 2
3 Build sprint 2 h 14 m left

Leaderboard Score

1 Bit Vectors 942
2 Null Pointers 918
3 Byte Club 901
4 Off by One 887

Live quiz · 1,204 playing0:08

Which of these sorting algorithms is stable?

Merge sort 64% Quick sort 21% Heap sort 15%

Game-based Pattern recall · level 6 38 s on this level

fest-portal.zipPortfolio review Code quality Design Documentation

Assessment Builder

Set your own questions

Superset Assessment Builder helps you assemble custom question papers from Superset’s question bank or your own content.

More than 20 question types
Question and option randomisation built in
AI evaluation on subjective questions
Configurable cut-offs and shortlist automation

The Assessment Builder assembling an SDE-1 coding round: live coding, technical MCQ, debugging and code-review sections with their marks, a coding problem open with its starter code and hidden tests, and the proctored schedule.

SDE-1 coding round · v2 120 min · 100 marks · N-of-M draw Draft

Sections

1 Live coding 2 problems · 40 marks
2 Technical MCQ 15 questions · 30 marks
3 Debugging 4 exercises · 20 marks
4 Code quality review 1 exercise · 10 marks
+ Add section

Question bank Yours and Superset’s

Problem 1 of 2 Medium

Merge overlapping intervals and return them sorted by start.

def merge(intervals):    # your code here    return []

8 test cases · 5 hidden 2 s · 256 MB · Python, Java, C++ and more

Arrays · Sorting

9:30 am · remote, proctored 1,200 candidates Proctored

Proctoring and integrity

AI proctoring ensures scores can be trusted

Superset layers lockdown, camera AI, AI-assistance detection and human review over every attempt.

Safe Exam Browser runs the test in a full-screen kiosk: task switching blocked, virtual machines and remote desktops refused, only allowlisted URLs reachable, and screen-share or AI tools auto-terminated before the first question renders. Where lockdown is not feasible, browser safeguards enforce full screen, count tab switches and monitor the clipboard.
Continuous face match against the verified identity, liveness prompts, and edge-AI object detection for phones, books and second faces. Dual proctoring pairs the candidate’s phone as a side-angle second camera, so the blind spot beside the laptop is on the record too.
Built for the AI era: detection of AI tool windows and hidden overlays, paste bursts and answer-latency signatures, keystroke-rhythm deviation from the candidate’s own baseline, and code stylometry that compares a submission’s style against the candidate and the cohort.
Recordings and watermarked snapshots land in an evidence vault behind an immutable event log. Every candidate report carries an integrity index alongside the marks, and flagged sessions queue for a human look.

Locked-down delivery settings: full-screen kiosk, virtual machines refused, a URL allowlist, background apps closed and browser safeguards, with a candidate's passed pre-test device check and a blocked tab switch.

Delivery settings SDE drive · Safe Exam Browser

Full-screen kiosk Task switching and shortcuts blocked
Refuse virtual machines And remote desktops, at launch
URL allowlist Only these destinations load docs.python.orgcppreference.com+2
Close background apps Screen share and AI tools terminated
Browser safeguards Fallback: full screen, tab-switch count, clipboard

Pre-test check Ananya Sharma · laptop

Camera
Microphone
No virtual machine
Screen share off
2 background apps closed

Ready to start

Tab switch blocked Karthik Rao · 10:14 · 1 of 3 allowed

Presence monitoring for one session: front camera with face match, the paired phone as a side camera, and a timeline of face match, an auto-pause, a phone flagged in frame and liveness prompts, with drive-wide counts above.

AI-assistance signals on one coding answer: a pasted block marked in the construction trace, a typing-rhythm break against the candidate's own baseline, and a card listing the signals behind the review, which a reviewer decides on.

Arjun Nair Coding round · question 2 of 2

Answer construction

1def merge(items):2    items.sort()3    out = []4    for s, e in items:5        if out and s <= out[-1][1]:6            out[-1][1] = max(out[-1][1], e)7        else:8            out.append([s, e])9    return out

Lines 4–8 pasted at 24:10

Typing rhythm

Own baseline This attempt

Why this answer is in review

Paste burst 5 lines inserted at 24:10
Rhythm break Outside the candidate’s own baseline for 2 min
Style shift Differs from the candidate’s earlier answers
Assistant windows None detected during the attempt

These are signals, not a verdict. A reviewer looks at the evidence and decides.

Open in review

The evidence pack for one candidate: a sealed event log, watermarked snapshots and the vaulted recording, the integrity index beside the marks, and a human review card with clear and uphold actions.

Karthik Rao SDE drive · report Integrity index 3 flags Marks 78 / 100

Event log · sealed

10:00:12 Identity verified
10:14:31 Tab switch blocked
10:31:05 Tab switch blocked
10:47:50 Phone in frame
11:02:18 Liveness prompt passed
11:30:00 Submitted

Snapshots

10:00:12

10:47:50

11:02:18

Camera and screen 90 min · in the evidence vault

Answer checks

Paste burstsNone
Typing rhythmOwn baseline
Clipboard1 paste logged

Human review Neha Joshi · 3 flags to check Clear flags Uphold flags

See the assessments on your own roles

A walkthrough with a hiring expert — pick a role you are hiring for and see the battery, the proctoring and the report it produces.

Schedule a product demo Talk to sales