The ParikshAI Platform

Deep learning meets academic rigor. Our proprietary AI models are trained exclusively on university-level examinations to provide human-level accuracy at machine speed.

app.pariksha.tech/evaluate/25421970
ParikshAI Enterprise
Evaluation Mode
Rishabh Galkar
Evaluator
RG
Scoring40.0/70(57.1%)
1
Q1
9/14
2
Q2
4.5/14
3
Q3
10.5/14
Part A
2.5
/3
or
Part A (OR)
/3
Part C
5
/7
or
Part C (OR)
/7
Part B
3
/4
or
Part B (OR)
/4
Optional note...
4
Q4
6.5/14
5
Q5
9.5/14
Submit Evaluation
25421970 · Network Theory|12 / 33⊕ 120%
DO NOT WRITE ANYTHING IN THIS PORTION

Q : 3

a) Initial Value Theorem.

By Initial Value Theorem :

lim f(t)  =  lim s · F(s)
t→0⁺s→∞

Given: F(s) = (2s + 1) / s(s + 2)

∴ lim s · F(s)

s→∞

= lim s(2s+1) / s(s+2)

= lim (2s+1) / (s+2)

=  2

T
AI Score83%
2.5/ 383%
+2.5m earned
AI Reasoning

Student correctly states the Initial Value Theorem and applies it to F(s) = (2s+1)/s(s+2). Limit computation and final value of 2 are accurate. Minor deduction for missing intermediate step notation.

Found on page: 12
app.ParikshAI.metaharbour.ai / rubrics / net-401
ParikshAI
Dashboard
Batches
Evaluators
Rubric Studio
Reports
Active Rubrics
NET 401 - Final
DC 302 - Midterm
EM 201 - Endsem
Rubric Studio

Network Theory · Semester Final

Criterion Name
Max Marks
Weight
Action
Concept Definition
Clear statement of the principle, theorem, or law being tested.
3 pts
25%
Formula & Derivation
Correct formula cited with complete step-by-step derivation shown.
4 pts
35%
Circuit Diagram
Accurately labeled circuit sketch or diagram with correct topology.
3 pts
25%
Units & Notation
Correct SI units and standard engineering notation used throughout.
2 pts
15%
Total Criteria: 4
Total Marks: 12Total Weight: 100%
Cognitive Mapping

Digitize your exact grading standards.

The engine doesn't just look for keywords. It understands context, derivations, and methodology. Translate your paper rubrics into a cognitive framework that the AI uses to evaluate students exactly how a human professor would.

  • Support for multi-part questions
  • Granular weighting by criteria
  • Custom negative marking rules
Massive Scale

Process 100,000 pages per hour.

Our AI engine is built for the intensity of final exam seasons. It ingests thousands of handwritten scripts, extracts the text using state-of-the-art OCR, and applies your rubric simultaneously across the entire batch.

  • 99.1% handwriting recognition accuracy
  • Under 2 seconds processing time per page
  • Identifies low-confidence anomalies instantly
app.pariksha.tech/admin/dashboard
P
ParikshAI
AI-Powered Evaluation
Main Menu
ParikshAI v1.0
Welcome to ParikshAI
AI-powered exam evaluation for universities
UNIVERSITIES
5
ONBOARDINGS
28
0 processing · 27 done
USERS
21
DOCUMENTS
134
Recent Onboardings
GUJARAT TECHNOLOGICAL UNIVERSITY – Bachelor of Engineering
Completed
GTU · Network Theory
1/1 (100%)
Gujarat Technological University – bachelors (Pharmacy)
Completed
GTU · Pharmaceutical Formulations
1/1 (100%)
Gujarat Technological University – bachelors (Pharmacy)
Completed
GTU · Pharmaceutical Formulations
1/1 (100%)
Quick Actions
AI Activity
27 completed
0 processing
134 docs
Compliance Rate 95%
app.pariksha.tech / evaluate / 45102
ParikshAI
Dashboard
Batches
Evaluators
Rubric Studio
Reports
Queue
6 scripts flagged for manual review
Script ID
#45102 · NET 401
Low Confidence: 71%

Q3. Two-Port Networks

"Z-parameters of the network: V₁ = Z₁₁I₁ + Z₁₂I₂. At port 1, I₂ = 0 ∴ Z₁₁ = V₁/I₁ = 10 Ω. By reciprocity Z₁₂ = Z₂₁ = 5 Ω..."

Missing SI units
AI Proposed Grade
4/ 7 marks
Rubric Breakdown
Concept Definition
3/3
Formula & Derivation
3/4
Circuit Diagram
2/3
Units & Notation
0/2
AI Feedback
Marked down for "Units & Notation" — SI units absent throughout. Please verify.
Human-in-the-Loop

You retain ultimate control.

The AI does the heavy lifting, but human evaluators maintain full authority. Any script that falls below the 98% confidence threshold is immediately flagged and routed to a human for manual review.

  • One-click overrides
  • Transparent audit trails for every mark
  • Beautiful review interface for professors
100k+
Pages processed per hour
99.1%
Handwriting OCR accuracy
< 2s
Processing time per page
99.8%
Grading accuracy vs. human

Built for the scale of Indian university examinations

ParikshAI is not a generic AI tool adapted for education. Every component — OCR engine, rubric evaluator, confidence scorer, and human review portal — is purpose-built for university-level written examinations in India.

Rubrics Studio — No-code grading criteria

Subject faculty define marking criteria, weightings, and model answers without any technical help. The Rubrics Studio translates paper marking schemes into a cognitive framework the AI uses to evaluate students exactly as a human professor would — including partial credit, alternative correct answers, and methodology-based marks.

AI Batch Engine — 100,000 pages per hour

ParikshAI's OCR engine achieves 99.1% accuracy on Indian university handwriting across English, Hindi, and regional scripts. The batch engine processes entire exam sets simultaneously — not sequentially — and assigns a confidence score to every answer. Uncertain answers are automatically held for human review.

Evaluator Portal — Human-in-the-loop

No grade is published without human approval. The Evaluator Portal surfaces only the scripts that need attention — typically 15–20% of a batch — with the original scan, AI score, and rubric side by side. Evaluators can override, approve, or send scripts back for re-evaluation. Every action is logged with a timestamp and evaluator ID.

Admin Dashboard — Full batch visibility

The administrator dashboard gives COE heads and examination controllers a live view of every active batch: scripts processed, scripts pending review, confidence distribution, and estimated completion time. Results can be locked and published by department once human review is complete.

Security and Data Residency

All student data is encrypted in transit using TLS 1.3 and at rest using AES-256. Data is stored exclusively on AWS infrastructure in the Mumbai (ap-south-1) region — no student data leaves India. ParikshAI operates as a data processor under the DPDP Act 2023 and provides a signed Data Processing Agreement to every institution partner.

LMS and ERP Integrations

ParikshAI connects to Moodle, Canvas, Blackboard Ultra, TCS iON, and Mettl via LTI 1.3, REST API, and webhooks. Grades write back directly to your gradebook. For proprietary SIS or ERP systems, our engineering team builds bespoke connectors — most integrations are live within four weeks.

FAQ

Frequently Asked Questions

What is ParikshAI and how does it work?
+
ParikshAI grades both handwritten scanned answer booklets and CBT responses at scale — against rubrics you define — assigning a confidence score to every answer. Low-confidence answers are automatically flagged and routed to human evaluators before grades are finalised. Result timelines drop from 2 months to under 2 weeks.
How accurate is the AI grading? Is it human-verified?
+
99.8% accuracy measured against human evaluator scores across 50,000+ reviewed booklets. No grade is published without passing through the Evaluator Portal. Every AI score carries a confidence percentage — scripts below your set threshold are mandatory-reviewed by a subject evaluator before results are locked.
How long does it take to evaluate 10,000 answer sheets?
+
End-to-end — AI grading, confidence flagging, and human review — under 2 weeks for 10,000 scripts. The AI engine runs at 2,400 scripts per hour. Manual grading of the same batch typically takes 2–3 weeks before human review even begins.
Is student answer data stored securely? Who owns it?
+
Your institution owns all data — scripts, rubrics, and grades. ParikshAI is a data processor: we never use your data for model training. Data is encrypted in transit (TLS 1.3) and at rest (AES-256), stored on Indian data centre infrastructure, and deleted on request. A signed Data Processing Agreement is provided to every partner.
What happens when the AI is unsure? How are disputes handled?
+
Answers below your confidence threshold cannot be published until a human evaluator approves them. For student disputes, ParikshAI provides a full per-script audit trail: the original scan, the rubric applied, AI score with reasoning, and every human action — satisfying university grievance redressal requirements.
Does it work with our existing scanning infrastructure?
+
Yes. ParikshAI accepts JPEG, PNG, TIFF, and multi-page PDF — standard output from MFD and document scanners already used by most examination departments. No specialist hardware required. For CBT, a REST API connects to Moodle, Mettl, and custom exam platforms.
Is ParikshAI compliant with UGC and AICTE guidelines?
+
ParikshAI is built with mandatory human-in-the-loop review, full audit trails, and institution-controlled grade locks — ensuring no AI score is published unilaterally. We work with each university's examination controller to map our workflow to their ordinance requirements. Compliance documentation is available for your academic council.
How do we set up rubrics? Do we need technical staff?
+
No technical staff needed. The Rubrics Studio is a no-code interface for subject faculty — define criteria, allocate marks, optionally upload a model answer — typically 15–30 minutes per question paper. Our onboarding team runs a rubric-building workshop with your department heads before the first batch.

Start Today

Ready to grade smarter and publish faster?

Start with 100 free scripts. No credit card, no lock-in.

Trusted by universities across Gujarat, Maharashtra, and Rajasthan.