AI-Driven EdTech App & Web Development
Budget: $3,000 – $5,000 USD
Project Brief – AI-Powered Education Modules (Phase 1)
Platform:
Mobile app (Flutter, Android & iOS)
Web-based dashboard (React / Vue / Angular — developer’s choice)
Backend: Node.js / Python with REST APIs
Database: PostgreSQL / MySQL (with offline sync layer)
Offline–Online sync capability for all modules
Scope Overview
We are building 5 integrated AI modules for an EdTech platform to enhance student learning, teacher productivity, and parental insights.
Modules to be developed in Phase 1:
Module A — Adaptive Learning Paths
Module B — AI Question & Exam Generator
Module C — Knowledge Gap Analysis
Module D — Automated Essay Grading
Module E — Mood Tracking (EMA)
Module Details
Module A — Adaptive Learning Paths (Personalised Recommendations)
Objective: Recommend a ranked sequence of topics/activities (notes, latihan tubi, assignments, recordings) that maximizes mastery and engagement.
Key Features:
Personalisation using IQ/EQ profiles, topic mastery, attendance, and engagement signals
Sequence model (DKT/SAKT) + contextual multi-armed bandit for activity selection
Spaced repetition scheduling
Explainable recommendations (top 3 reasons displayed to student)
Inputs: IQ/EQ data, mastery estimates, practice performance, class timetable, attendance logs, streaks, device constraints.
Outputs: JSON array of top N recommendations with {topicId, activityType, expectedGain, confidence, reasons[]}
APIs:
GET /ai/adaptive/recs?userId=&subjectId=&n=10
POST /ai/adaptive/feedback
Offline/Sync: Cache next 20 recs; queue feedback and reconcile on reconnect.
KPIs: +X% mastery gain, reduced time-to-mastery, higher retention.
Module B — AI Question & Exam Generator
Objective: Generate syllabus-aligned questions & mock exams in BM/EN with difficulty control.
Key Features:
Template-driven MCQs/short answers with distractor banks
LLM-assisted drafting with quality filters (plagiarism, hallucination, bias scan)
IRT calibration from pilot responses
Teacher UI for human-in-loop review
Outputs:
Item: {stem, options, answerKey, rationale, difficulty, tags, lang}
Paper: {items[], blueprint, timing, markingScheme}
APIs:
POST /ai/qgen/item
POST /ai/qgen/paper
POST /ai/qgen/validate
Offline/Sync: Cache generated sets; sync metadata later.
KPIs: Teacher acceptance rate, <1% item error rate, predictive validity for difficulty.
Module C — Knowledge Gap Analysis
Objective: Identify weak skills per student and suggest improvement sequence.
Key Features:
Bayesian Mastery Model + Topic Knowledge Graph
Diagnostic mini-tests for low-confidence areas
Class heatmap for teachers; simplified summaries for parents
Outputs: {topicId, mastery, confidence, blockers[], nextBest[]}
APIs:
GET /ai/gaps/student/{id}?subjectId=
GET /ai/gaps/class/{classId}
Offline/Sync: Store snapshots for dashboard; merge on sync.
KPIs: Correlation with future performance; reduced reteach time.
Module D — Automated Essay Grading (BM & EN)
Objective: Score essays (100–600 words) with rubric-based subscores + feedback.
Key Features:
Pre-checks for plagiarism, AI-generated likelihood, length, language ID
Fine-tuned transformer for scoring: content, coherence, grammar, vocab, mechanics
Teacher override + calibration feedback loop
Outputs: {overall, subscores{}, feedback[], confidence, flags{}}
APIs:
POST /ai/essay/score
POST /ai/essay/explanations
Offline/Sync: Cache last 10 submissions; sync later.
KPIs: ≥ 0.75 quadratic weighted kappa vs human grading, <5s processing time.
Module E — Mood Tracking (EMA)
Objective: Capture mood/energy/stress signals to personalise learning support.
Key Features:
Opt-in emoji-based mood survey 2–3×/week
Passive engagement signals (optional)
Aggregates for parents/teachers without raw journal text
Outputs: {mood, energy, stress, context, timestamp}
APIs:
POST /ai/mood/entry
GET /ai/mood/summary?window=30d
Offline/Sync: Queue entries and sync by server time.
KPIs: Completion rate, correlation with engagement dips.
General Requirements
Authentication & Roles: Student, Teacher, Parent, Admin
UI/UX: Mobile-first design, simple navigation for students
Data Privacy: PDPA-compliant storage; encrypted in transit & at rest
Explainability: All AI outputs should include reasoning where applicable
Offline Support: All modules must work offline and sync later
Scalability: Backend should handle 100,000–500,000 concurrent live users for live classes
Deliverables
Fully functional API layer for each module
Mobile app integration (Flutter) for Student/Parent/Teacher dashboards
Web dashboard for teachers/admins
Documentation: API specs, deployment steps, and user manuals
Testing: Unit tests + integration tests for all modules
Deployment on AWS (provided account)
Timeline
Week 1–2: Architecture setup, DB schema, and API stubs
Week 3–6: Module A & B core logic + integration
Week 7–9: Module C & D
Week 10: Module E
Week 11: Testing, bug fixes, documentation
Week 12: Final deployment
Platform:
Mobile app (Flutter, Android & iOS)
Web-based dashboard (React / Vue / Angular — developer’s choice)
Backend: Node.js / Python with REST APIs
Database: PostgreSQL / MySQL (with offline sync layer)
Offline–Online sync capability for all modules
Scope Overview
We are building 5 integrated AI modules for an EdTech platform to enhance student learning, teacher productivity, and parental insights.
Modules to be developed in Phase 1:
Module A — Adaptive Learning Paths
Module B — AI Question & Exam Generator
Module C — Knowledge Gap Analysis
Module D — Automated Essay Grading
Module E — Mood Tracking (EMA)
Module Details
Module A — Adaptive Learning Paths (Personalised Recommendations)
Objective: Recommend a ranked sequence of topics/activities (notes, latihan tubi, assignments, recordings) that maximizes mastery and engagement.
Key Features:
Personalisation using IQ/EQ profiles, topic mastery, attendance, and engagement signals
Sequence model (DKT/SAKT) + contextual multi-armed bandit for activity selection
Spaced repetition scheduling
Explainable recommendations (top 3 reasons displayed to student)
Inputs: IQ/EQ data, mastery estimates, practice performance, class timetable, attendance logs, streaks, device constraints.
Outputs: JSON array of top N recommendations with {topicId, activityType, expectedGain, confidence, reasons[]}
APIs:
GET /ai/adaptive/recs?userId=&subjectId=&n=10
POST /ai/adaptive/feedback
Offline/Sync: Cache next 20 recs; queue feedback and reconcile on reconnect.
KPIs: +X% mastery gain, reduced time-to-mastery, higher retention.
Module B — AI Question & Exam Generator
Objective: Generate syllabus-aligned questions & mock exams in BM/EN with difficulty control.
Key Features:
Template-driven MCQs/short answers with distractor banks
LLM-assisted drafting with quality filters (plagiarism, hallucination, bias scan)
IRT calibration from pilot responses
Teacher UI for human-in-loop review
Outputs:
Item: {stem, options, answerKey, rationale, difficulty, tags, lang}
Paper: {items[], blueprint, timing, markingScheme}
APIs:
POST /ai/qgen/item
POST /ai/qgen/paper
POST /ai/qgen/validate
Offline/Sync: Cache generated sets; sync metadata later.
KPIs: Teacher acceptance rate, <1% item error rate, predictive validity for difficulty.
Module C — Knowledge Gap Analysis
Objective: Identify weak skills per student and suggest improvement sequence.
Key Features:
Bayesian Mastery Model + Topic Knowledge Graph
Diagnostic mini-tests for low-confidence areas
Class heatmap for teachers; simplified summaries for parents
Outputs: {topicId, mastery, confidence, blockers[], nextBest[]}
APIs:
GET /ai/gaps/student/{id}?subjectId=
GET /ai/gaps/class/{classId}
Offline/Sync: Store snapshots for dashboard; merge on sync.
KPIs: Correlation with future performance; reduced reteach time.
Module D — Automated Essay Grading (BM & EN)
Objective: Score essays (100–600 words) with rubric-based subscores + feedback.
Key Features:
Pre-checks for plagiarism, AI-generated likelihood, length, language ID
Fine-tuned transformer for scoring: content, coherence, grammar, vocab, mechanics
Teacher override + calibration feedback loop
Outputs: {overall, subscores{}, feedback[], confidence, flags{}}
APIs:
POST /ai/essay/score
POST /ai/essay/explanations
Offline/Sync: Cache last 10 submissions; sync later.
KPIs: ≥ 0.75 quadratic weighted kappa vs human grading, <5s processing time.
Module E — Mood Tracking (EMA)
Objective: Capture mood/energy/stress signals to personalise learning support.
Key Features:
Opt-in emoji-based mood survey 2–3×/week
Passive engagement signals (optional)
Aggregates for parents/teachers without raw journal text
Outputs: {mood, energy, stress, context, timestamp}
APIs:
POST /ai/mood/entry
GET /ai/mood/summary?window=30d
Offline/Sync: Queue entries and sync by server time.
KPIs: Completion rate, correlation with engagement dips.
General Requirements
Authentication & Roles: Student, Teacher, Parent, Admin
UI/UX: Mobile-first design, simple navigation for students
Data Privacy: PDPA-compliant storage; encrypted in transit & at rest
Explainability: All AI outputs should include reasoning where applicable
Offline Support: All modules must work offline and sync later
Scalability: Backend should handle 100,000–500,000 concurrent live users for live classes
Deliverables
Fully functional API layer for each module
Mobile app integration (Flutter) for Student/Parent/Teacher dashboards
Web dashboard for teachers/admins
Documentation: API specs, deployment steps, and user manuals
Testing: Unit tests + integration tests for all modules
Deployment on AWS (provided account)
Timeline
Week 1–2: Architecture setup, DB schema, and API stubs
Week 3–6: Module A & B core logic + integration
Week 7–9: Module C & D
Week 10: Module E
Week 11: Testing, bug fixes, documentation
Week 12: Final deployment