PNACADEMY
COSE v10.0 · Neural Terminal
Trilingual · Forensic · Secure
Accent Color
Step 01 — Face Biometric

AWAITING CAMERA...

Step 02 — Select Your Role
Step 03 — Credentials
ID
PW
Credentials are provisioned by your Super User. First-time / temporary passwords must be changed immediately after login.
PNACADEMY
COSE v10.0 · Select Your Role
Access is scoped to your responsibilities in the evaluation pipeline
TC

Teacher / Evaluator

Review AI-evaluated scripts assigned to your subject, confirm or adjust marks, and see the full reasoning behind every score.

  • View assigned scripts queue
  • Review AI evaluation & mark breakdown
  • Confirm / modify marks with audit log
  • Full explanation for every mark
ENTER TEACHER DASHBOARD →
EM

PNAcademy Employee

Run the evaluation pipeline: upload the teacher's solution once, bulk-upload student sheets, and auto-route them to the right evaluator.

  • Bulk upload student answer sheets
  • Upload teacher solution key (once)
  • Auto-assign scripts by subject expertise
  • Run AI evaluation across the batch
ENTER EMPLOYEE DASHBOARD →
EC

AKU Exam Controller

Oversee the full evaluation pipeline across all colleges — view results, analytics, and approve or override final marks.

  • View all evaluated results
  • Analytics & pass/fail reports
  • Override marks with justification
  • Final approval & result publishing
ENTER CONTROLLER DASHBOARD →
Provision New Account
Create User
Assign a role, an ID (User ID for teachers, Employee ID, or University/Controller ID), and a temporary password. The user will be forced to change it on first login.
Role
ID (User ID / Employee ID / Controller ID)
Display Name
Temporary Password
Bulk Provisioning
Bulk Upload — Create Multiple Accounts at Once
Upload a CSV or Excel file to automatically provision many Teacher / Employee / Controller accounts in one go. Expected columns: Role, ID, Name, Password.
Header row optional — Role (teacher/employee/controller), ID, Name, Password ⬇ Download CSV Template
All Accounts
Reset a temporary password or deactivate an account.
IDNameRoleStatusActions
LIVE AUDIT ACTIVE
00:00
Workflow Guide
The 5-Step Employee Dashboard Workflow
Follow these five steps in order to take a batch of exams from raw student lists to fully AI-graded, teacher-reviewed scripts. Every step has a Manual Fallback if the automated path can't be used.
1
Bulk Onboard Users
Students & Teachers
Upload Separate Lists via CSV/Excel. Missing names or duplicate roll numbers are highlighted in red for instant on-screen fixes. Manual single-record entry always available.
📤 Upload Student List (CSV/Excel) · 📤 Upload Teacher List (CSV/Excel)
→
2
Upload University Master Key
Answers & Marking Schema
Upload the official answer key (PDF, Word, or Image). The system auto-extracts answers, max marks per question, and the full Marking Schema. Manual Entry (paste text) always available.
📄 Upload Master Solution & Marking Schema
→
3
Bulk Upload Answer Sheets
Student Answer Sheets
Drag & drop a whole folder of scanned Student Answer Sheets. Files land in Azure Blob Storage, which automatically triggers the Service Bus FIFO queue. Manual single-sheet upload always available.
📁 Drag & Drop Answer Sheets Here (Bulk Upload)
→
4
Automated Batch Processing
Digitize, Auto-Map & Save
Azure Durable Functions automatically pick up each file from the Service Bus queue in strict order, run it through the High-Fidelity Multimodal OCR Pipeline, extract the Roll Number/Barcode, and save the result straight to the database — no manual click needed. Torn or unreadable codes pause for ⚠ Manual Update Required.
⚡ AUTOMATED — BLOB TRIGGER → SERVICE BUS → DURABLE FUNCTION → DB
→
5
Run AI Evaluation
Grade & Route to Teacher
The AI compares digitized sheets against the Master Key, applies marks per the Marking Schema, then forwards graded scripts to the assigned Teacher/Evaluator for final manual review.
▶ RUN AI EVALUATION ON PIPELINE
Pipeline Overview
Batch Evaluation Pipeline
Every script is hashed against the teacher key — identical content always yields identical, cached marks.
0
TOTAL SCRIPTS
0
PENDING AI
0
AI EVALUATED
0
TEACHER REVIEWED
0
SERVED FROM CACHE
Step 02
Upload the University Master Key — Master Solution & Marking Schema
Uploaded once per subject — every student script in that subject is compared against this single master key. Manual Entry is always available: paste the answer key text directly instead of uploading a file.
Subject
Master Solution / Answer Key Text
EFFECTIVE MARKING SCHEME (Filter OR Default) — this is what the AI uses in Step 05
Q. No.
Marks
Q. No.Filter MarksDefault (Master Key)Effective Marks Used
Manual Fallback
Add a Single Student or Teacher (Evaluator Roster)
No CSV/Excel handy? Onboard one record at a time. Scripts auto-route to the teacher registered as an expert for that subject.
✎ MANUAL ADD — TEACHER
✎ MANUAL ADD — STUDENT
Forensic Digitization Engine
High-Fidelity Multimodal OCR Pipeline
Scan a physical/handwritten answer sheet before adding it to the batch — verbatim transcription · layout preservation · diagram mapping · LaTeX equation extraction.
VERBATIM LAYOUT PRESERVE DIAGRAM MAP
ACQ
Image Acquire & Pre-Process
LAY
Layout Analysis
SEG
Line & Word Segment
SCR
Script Identification
OCR
Feature Extract & OCR
SEM
Post-Process & Semantic Fix
STUDENT SHEET UPLOAD (IMAGE OR PDF)
📄
Drop image or PDF, or click to upload
Handwritten sheets · Printed answer booklets · Scanned PDFs
Devanagari · Latin · Perso-Arabic script
JPGPNGWEBPPDF
filename.jpg
OCR STATUS LOG
[READY] High-Fidelity OCR Engine initialized.
[INFO] Mode: VERBATIM — no autocorrect active
[WAIT] Upload a scanned sheet to begin...
0
CHARS EXTRACTED
0%
CONFIDENCE
WORD-FOR-WORD TRANSCRIPTION — NO CORRECTION APPLIED
[Output will appear here...]
[Diagram relational mapping will appear here if diagrams are detected...]
EQUATIONS DETECTED AND CONVERTED TO LaTeX
[LaTeX output will appear here if math equations are detected...]
PAGE STRUCTURE & LAYOUT ANALYSIS
[Layout analysis will appear here...]
Step 03 · High-Fidelity Multimodal OCR Pipeline
Bulk Upload Student Answer Sheets
Bring in the raw exam papers. Drop a folder containing hundreds of scanned Student Answer Sheets — image, PDF, Word, PowerPoint, Excel, text, or virtually any format. Uploads land directly in Azure Blob Storage, which triggers an Azure Service Bus Queue (FIFO, sessions enabled); Azure Durable Functions then pick up each file sequentially and run it through the High-Fidelity Multimodal OCR Pipeline (verbatim transcription · layout preservation · diagram mapping · LaTeX equation extraction), extract the Roll Number/Barcode identity, and save the mapped result straight into the database — automatically, in order, with no manual step required. Prefer one sheet at a time? Use the Manual Single-Sheet Upload above instead.
HIGH-FIDELITY OCR QR/BARCODE MATCH ONE-BY-ONE PIPELINE
ACQ
Image/File Acquire & Pre-Process
LAY
Layout Analysis
SEG
Line & Word Segment
SCR
Script Identification
OCR
Feature Extract & OCR
QR
QR/Barcode Identity Match
⬡ VERBATIM TRANSCRIPTION · LAYOUT PRESERVATION · DIAGRAM MAPPING · LATEX EQUATION EXTRACTION — runs automatically per file below
Default Subject for this Batch
📁
📁 Drag & Drop Answer Sheets Here (Bulk Upload)
Image · PDF · Word · PowerPoint · Excel · Text · any scanned format — lands in Azure Blob Storage and processes automatically, one file at a time, strict FIFO. Or click to select multiple files manually.
JPGPNGWEBPHEICTIFFPDFDOCDOCXPPTPPTXXLSXLSXTXTRTFODTANY
Step 01
Bulk Onboard Users (Students & Teachers)
Before processing exams, the system needs to know who is taking them and who is grading them. Two clear, Separate Lists keep student and teacher databases independent.
STUDENT DATABASE
Columns expected: name, roll, college, subject (header row optional). Missing names or duplicate roll numbers are highlighted in red for on-screen fixes.
TEACHER DATABASE
Columns expected: name, roll/staff ID, college, subject expertise (header row optional). Duplicates are flagged automatically.
Manual Fallback
Manual Entry — Student Answer Sheets
No file to upload, or a script needs typing in by hand? Add each script's details manually here, then submit the batch — subject auto-routes each script to its expert teacher.
Step 05
Run AI Evaluation
Grade the digitized, organized papers. The AI evaluates every question in the script — no upper limit on question count or total marks — strictly against the Effective Marking Scheme (Marking Schema Filter OR the default marks extracted from the Master Key), then forwards graded scripts to the assigned Teacher/Evaluator's dashboard for final manual review. Deterministic (temperature=0, content-hash cached over the script, master key, AND the effective marking scheme) — same script + same key + same marking scheme always produces the exact same marks, even after a refresh or re-run.
⬡ DETERMINISTIC SCORING ACTIVE
ScriptStudentSubjectAssigned TeacherStatusAI ScoreIntegrity CheckActions
Session Identity
—
Evaluator Telemetry & Analytics — live stats for your account only.
0
ASSIGNED TO ME
0
AWAITING MY REVIEW
0
REVIEWED & SUBMITTED
—
AVG TIME / COPY
0
TAB/FOCUS SWITCHES (SESSION)
Evaluator Telemetry
Performance & Proctoring Analytics
Azure Application Insights / Monitor telemetry for the currently logged-in evaluator — manual evaluation throughput, time-per-copy trend, and integrity signals.
TIME PER COPY (LAST 10 EVALUATED)
COMPLETION RATE
MOST RECENTLY EVALUATED COPY
No copies evaluated yet this session.
Queue
Scripts Assigned To You
AI has pre-evaluated each script deterministically. Open one to confirm, adjust, and see the full reasoning behind every mark.
ScriptStudentSubjectStatusAI ScoreFinal ScoreTime Spent
University Overview
Result Pipeline Summary
0
TOTAL SCRIPTS
0
APPROVED
0
PENDING REVIEW
0
AVG SCORE %
Analytics
AI vs Teacher-Confirmed Marks
Divergence between AI-suggested and teacher-confirmed marks across all reviewed scripts.
Institutional Analytics
Exam Health Dashboard
Predictive insights for Registrars & Deans — question difficulty heatmap and evaluator calibration tracking.
QUESTION DIFFICULTY HEATMAP

Flags questions with high failure/low-average rates across all evaluated scripts.

EVALUATOR CALIBRATION TRACKING

Monitors strict vs. lenient grading trends — average teacher deviation from AI-suggested marks.

Analytics Grid
College → Subject → Teacher Breakdown
Copies checked and marks awarded, drillable by College, Subject, and Teacher.
CollegeSubjectTeacherCopies CheckedTotal Marks AwardedAvg Marks / Copy
Publishing Engine
Bulk-Publish Results
Filter by Academic Year, College, or Subject, then publish — this locks in approved marks and triggers the automated AI report workflow for every matching student.
Academic Year
College
Subject

Only scripts with status Teacher Reviewed or Approved that match the selected filters will be published. Publishing is irreversible for a script once its status becomes Published.

Export & Edit
Download Results / Secure Edit

Downloads include Roll Number, Student Name, College Name, Subject, and Marks. Edit Mode allows inline correction of Final Score directly in the results table below — every change is written to the tamper-proof audit log with evaluator ID, timestamp, and before/after values.

All Results
Evaluated Scripts — View, Override & Approve
ScriptStudentCollegeSubjectTeacherAI ScoreFinal ScoreStatus
Security & Proctoring
Granular Security Policy Engine
Manage anti-cheat measures at the Teacher level or the College level. Toggle any parameter, or add a custom restriction not covered below.
Apply Scope
Add Custom Restriction
0 / 0 ACTIVE FOR THIS SCOPE
Live Intelligence Dashboard
Triple-Tier Evaluation Monitor
EV
Evaluator Tier
Total
0
Assigned
Checked
0
Evaluated
Pending
0
Awaiting
Live
0
In Progress
CL
College Tier
Total
0
College Sheets
Checked
0
Done
Pending
0
Queue
Live
0
Active
UN
University Tier
Total
0
University Sheets
Checked
0
Complete
Pending
0
Queue
Live
0
Active
Session Metadata
Evaluator Name
Student Full Name
Roll Number
College / Institution
Zero-Distraction Workspace
🔒 Script & Master Key — Locked for Evaluation
Digitization, OCR, and answer-sheet ingestion are handled entirely by PNAcademy Operations before a script reaches you. The student script and Master Answer Key below are read-only for the duration of this session — this keeps your workspace focused purely on grading, with no technical overhead.
IMMUTABLE TAMPER-EVIDENT
Need to fix a script or key? A senior admin can grant a temporary unlock — every unlock is written to a tamper-proof audit log.
Master Key (Teacher) 🔒 LOCKED

Ingested and verified by Operations · read-only

Student Answer Sheet 🔒 LOCKED

Digitized by Operations · read-only

OR
Smart OR-Question Logic Active: If your question paper has "Q1 OR Q2" type choices, the AI will automatically detect both answers, select the best one, and eliminate duplicate marks. Mark OR questions clearly in the Master Key.
Engineering Blueprint
60-Feature Architecture — All Phases Active
P1 · 15 features P2 · 15 features P3 · 15 features P4 · 15 features
P1

PHASE 1 — Advanced AI Handwriting & Logic

Forensic ICR · Literal Overlay · Semantic Mapping · Step-Wise Scoring

15
▼
F-01
Forensic ICR Engine
Recognizes cursive and messy handwriting by analyzing stroke patterns rather than just character shapes — achieving near-human accuracy on poor-quality scripts.
F-02
Literal Transcription Overlay
Displays digitized text directly over the handwriting image so the teacher can verify OCR accuracy instantly with a side-by-side comparison.
F-03
"Best of N" Choice Resolver
Grades all 8 attempts when only 6 are required; automatically highlights the top 6 scoring answers, eliminating double-counting.
F-04
Hinglish / Code-Switching Support
Seamlessly transitions between Hindi and English scripts within a single sentence — handles mid-sentence script switches without accuracy loss.
F-05
Diagram Geometry Validation
AI checks if a hand-drawn diagram contains all mandatory labeled components, awarding partial marks for partially correct diagrams.
F-06
Formula-to-LaTeX Conversion
Transforms handwritten mathematical equations into clean, editable digital LaTeX format — enabling precise symbolic comparison against the master key.
F-07
Step-Wise Partial Scoring
Awards marks for correct methodology in long-form answers even if the final result is wrong. Each logical step evaluated independently.
F-08
Handwriting Sentiment Analysis
Detects "Confidence Levels" by analyzing pen pressure and strike-through patterns. High strike-through density may indicate uncertainty or knowledge gaps.
F-09
Semantic Synonym Mapping
Maps equivalent meanings across languages — Water = H₂O = Paani. Cross-language semantic matching ensures students writing in Hindi/Urdu are graded fairly.
F-10
Contextual Tokenization
Differentiates ambiguous characters — '5' vs 'S' or '1' vs 'I' — based on subject context (Math vs. English), preventing OCR misreads.
F-11
Smart "OR" Logic Resolver
Prevents double-marking by identifying when Q1 and Q2 are optional choices. Grades both and selects the higher-scoring answer.
F-12
Keyword Heatmapping
Highlights essential technical terms within the handwritten scan, providing the teacher with instant visual feedback on keyword coverage.
F-13
Dynamic Rubric Calibration
Suggests score adjustments if marking pattern deviates significantly from the batch average — detecting both harsh and lenient evaluators.
F-14
Reference Answer Expansion
Provides the teacher with 5 different accepted correct answer variations for each question, reducing risk of penalizing valid alternative approaches.
F-15
Redaction Engine (Blind Marking)
Automatically masks student names and photos before presenting the answer sheet to the evaluator — ensuring 100% blind marking.
P2

PHASE 2 — Hyper-Security & Forensic Proctoring

Iron Dome · Biometrics · Gaze Tracking · Audio Forensics · Screenshot Block

15
▼
F-16
Multi-Person Notification
Instant alert triggered when more than one face is detected in the camera view during evaluation — logs the incident automatically.
F-17
Suspicious Object Detection
Real-time AI detection of smartphones, tablets, or unauthorized books in the camera frame — triggers alert and logs the incident automatically.
F-18
Evaluator Gaze Tracking
Uses webcam to ensure the teacher's eyes remain on the script — detects prolonged gaze-away events that may indicate phone consultation.
F-19
Audio Forensic Audit
Detects background whispers or dictation audio during the marking process — identifies potential AI-assisted answer generation via voice input.
F-20
Keystroke Biometrics
Identifies the evaluator by their unique typing rhythm and cadence — prevents account sharing and ensures only the assigned teacher evaluates.
F-21
Face-Match Liveness Detection
Requires the evaluator to perform a blink or movement action — proving a live human is present, not a static photo in front of the camera.
F-22
"Iron Dome" Lockdown
Blocks screenshots, tab-switching, and screen recording on the device during marking sessions — hardware-level secure evaluation environment.
F-23
IP-Geofencing
Restricts marking tasks to authorized University GPS coordinates — evaluator attempting remote marking outside campus is automatically locked out.
F-24
Forensic Invisible Watermarking
Embeds invisible tracking data in on-screen rendering — any leaked screenshot can be traced back to the specific evaluator session.
F-25
Screen Mirroring Block
Disables the application if it detects the screen is being cast to a secondary monitor — preventing collaborative cheating sessions.
F-26
AI Assistant Detection
Blocks external AI tools (ChatGPT, Gemini, etc.) from being used to auto-generate feedback or marks during evaluation.
F-27
Session Heartbeat Monitor
Auto-locks the interface if the evaluator leaves their seat for more than 30 seconds — prevents unattended sessions.
F-28
Incident Evidence Capture
Records a 10-second video snippet whenever a mark is manually changed — creating an auditable forensic log of every mark modification.
F-29
Peripheral Lock
Ensures no unauthorized USB or recording devices are active on the system — prevents capture devices from intercepting answer sheet data.
F-30
Shadow Detection
Alerts if a second face is partially visible in the camera periphery — detecting over-shoulder surveillance even when only partially in frame.
P3

PHASE 3 — Deep Analytics & Pedagogical Insights

Cognitive Mapping · Fatigue Analytics · Predictive Modeling · Auto-Tutoring

15
▼
F-31
20,000+ Word Content Depth
A massive semantic engine enabling high-accuracy subjective marking — evaluates conceptual depth, logical progression, and vocabulary at scale.
F-32
T20 Manhattan Charts
Match-style bar graphs comparing AI marks vs. Teacher marks across 5 performance dimensions on every completed audit report.
F-33
Cognitive Gap Mapping
Reports whether a student is failing due to Fact Errors (wrong information) or Logic Errors (flawed reasoning) — enabling targeted remediation.
F-34
Evaluator Fatigue Analytics
Flags "Speed-Marking" patterns when a teacher evaluates papers abnormally fast — ensuring every paper receives adequate evaluation time.
F-35
Regional Heatmaps
Maps district-wise performance trends for University administrators — enabling data-driven identification of underperforming regions.
F-36
Predictive Pass/Fail Modeling
Projects final batch results in real-time as evaluations are completed — allowing administrators to prepare remedial support before results are published.
F-37
6-Week Deep Roadmap
Generates a 1,000+ word personalized study plan for every student — structured week-by-week with specific topics, resources, and practice exercises.
F-38
5,000+ Word Topic Analysis
Detailed per-student report of strengths and weaknesses across each topic covered in the exam.
F-39
Live Peer-Benchmarking
Shows the teacher the state-wide average for the current question being evaluated — providing real-time context to prevent grading drift.
F-40
Plagiarism Pattern Matcher
Identifies mass-cheating events by detecting when multiple papers contain statistically identical text blocks — distinguishes group fraud from individual copying.
F-41
Voice-Annotated Feedback
Allows teachers to record audio comments directly on the student's answer sheet — richer feedback without additional typing time.
F-42
Sentiment-to-Report Sync
Includes the Handwriting Confidence Score in the final student report — giving educators insight into psychometric dimensions beyond content accuracy.
F-43
Re-evaluation Difference Report
Explains specifically why a second evaluator changed marks — generates a structured justification report highlighting re-interpreted criteria.
F-44
Historical Growth Tracking
Compares current exam performance to previous semesters — visualizing longitudinal academic growth and identifying students requiring intervention.
F-45
Auto-Tutoring Guide
Generates a curated list of recommended videos, textbook chapters, and practice problems targeted at the student's specific weaknesses.
P4

PHASE 4 — University Logistics & Scale

100K Concurrency · WORM Storage · Offline Sync · Full Audit Log

15
▼
F-46
Triple-Tier Dashboard
Separate purpose-built views for Evaluator, College, and University roles — each with context-appropriate metrics and data scoping.
F-47
One-Click Re-evaluation Portal
Direct student-facing interface for challenging marks — submits a formal re-evaluation request triggering blind double-evaluation automatically.
F-48
Automated Summation Audit
100% accurate mathematical totalling of all question marks — eliminates human addition errors that historically cost students marks.
F-49
Blind-Double Evaluation
Two teachers independently grade one paper — system flags discrepancies greater than 10% for third-party review without revealing either evaluator's marks.
F-50
Digital Ink Overlay
Full stylus support for circling, underlining, and annotating directly on the digital script — replicating the natural pen-and-paper marking experience on tablets.
F-51
Multi-Device State Sync
Seamlessly switch from PC to tablet mid-session without losing proctoring data, mark history, or session state.
F-52
Archival WORM Storage
Write-Once Read-Many secure storage — once finalized, no record can be altered or deleted, providing legally defensible archive integrity.
F-53
Metadata Headers
Hardcodes Student Name, Roll Number, and Evaluator ID into every generated PDF — ensures chain-of-custody documentation for every script.
F-54
High-Concurrency Engine
Built to handle 100,000+ simultaneous users without performance degradation — supports university-wide exam season with zero downtime SLA.
F-55
Offline-Sync Mode
Allows full marking without active internet — locally caches all work and syncs to the server automatically once connection is restored.
F-56
Evaluator Sandbox
A training mode for new teachers to practice university evaluation standards on sample papers before live evaluation begins.
F-57
Dynamic ID Watermarking
Overlays the teacher's name and ID visually across the on-screen script — psychological deterrent against photographing and leaking answer sheets.
F-58
Full Activity Audit Log
Records every mouse click, mark change, and navigation event with timestamps — complete forensic trail for dispute resolution and compliance.
F-59
Multi-Lingual Admin UI
Dashboard available in Hindi and English — ensuring regional university staff can navigate administrative controls without language barriers.
F-60
One-Click Result API
Direct API push to university portals for instant result publication — eliminates manual data entry, results appear on student portals within seconds.
Human Moderation
Audit Log & Mark Moderation — Question-by-Question
AI suggested marks are editable. Every question submitted is evaluated and listed below, no upper limit. All changes are forensically logged.
0 QUESTIONS
LIVE PROCTORING ACTIVE: Video & Audio captured for this moderation session.
Question / CriteriaAI SuggestedTeacher MarksReason & Audit Log
0
MODERATED PERFORMANCE SCORE
Evaluator
—
Student
—
Roll No.
—
College
—
Script
—

Point-by-Point Solution Comparison

TEACHER MODEL SOLUTION
STUDENT SUBMISSION

Topic-Wise Strengths (5000+ Words)

Topic-Wise Weak Points (5000+ Words)

6-Week Improvement Roadmap (1000+ Words)

Extensive Analytic Audit Summary