AI-AugmentedOpen SourceNTA Level 5/6
Browser‑based educational simulation platform for CS students – Arusha Technical College, Tanzania. Built by 7 students (Core‑First Engineering).
Interactive summary – jump to section

1. Executive summary

What: ATC SimLab delivers interactive, AI‑powered computer science simulations (sorting, networking, memory) for NTA Level 5/6. Students manipulate algorithms, ask Mistral‑powered tutor, and receive auto‑generated quizzes. Cost: ~$0.80 / semester for 200 students (Mistral small).

27% avg learning gain
based on PhET Rwanda study
Offline‑first PWA
works on 3G / low‑end devices
200+ ATC students
target pilot cohort
7 roles
student‑led agile team
mindmap root((ATC SimLab)) Problem No NTA‑aligned sims Abstract CS hard to grasp Expensive Western tools Solution Sorting Visualiser (5 sorts) Network Topology (Star/Ring/Mesh) Memory/Stack visualiser AI Tutor + Quiz Generator Tech BOOM BOOM + PixiJS/Canvas FastAPI + Supabase Mistral AI + Codestral Cloudflare + Render Impact 1,000,000 students Year 1 Open source MIT Expand to UDOM/DIT VETA/TVET recognition
📌 MVP exit criteria (Week 20): 3 simulations, AI tutor, quiz generator, 50+ students, 1 lecturer assigned, handover docs complete.

2. Competitive landscape analysis

ATC SimLab benchmarks against 16 platforms (PhET, uCertify, Labster, Virtual Labs India, etc.). Unique position: high CS specificity + low cost/open source + NTA alignment.

quadrantChart title Platform positioning (CS specificity vs openness) x-axis "Low CS specificity" --> "High CS specificity" y-axis "High Cost / Closed" --> "Low Cost / Open" quadrant-1 "Open niche" quadrant-2 "Open general" quadrant-3 "Closed general" quadrant-4 "Closed niche" "PhET": [0.2, 0.85] "uCertify": [0.8, 0.2] "Labster": [0.15, 0.25] "Virtual Labs India": [0.75, 0.7] "ATC SimLab": [0.92, 0.95]
Table 1: Key differences – ATC SimLab vs. alternatives
PlatformCS‑specificAI tutorLow bandwidthOpen sourceSuitable for ATC
PhETPartialNo (not CS)
uCertifyCost prohibitive
Virtual Labs IndiaAging tech, no AI
ATC SimLab✅✅✅ Mistral✅ PWA✅ MITBuilt for NTA

3. 20‑week roadmap & interactive milestones

gantt title ATC SimLab MVP Timeline (20 weeks) dateFormat YYYY-MM-DD axisFormat %b %d section Phase 0: Foundation GitHub + Supabase + FastAPI + React scaffold :p0, 2026-01-08, 21d section Phase 1: Sorting Sim MVP Canvas + Bubble Sort + 5 sorts + AI Tutor + Quiz :p1, after p0, 49d Student validation (20 users) :milestone, m1, after p1, 3d section Phase 2: Network + Memory Network Topology, Memory/Stack, AI Narrator :p2, after m1, 42d Formal assessment (40 students) :milestone, m2, after p2, 5d section Phase 3: Institutionalise Dept presentation, lecturer onboarding, handover :p3, after m2, 28d MVP complete :milestone, m3, after p3, 1d
Milestone M4 (MVP complete)
Week 20 – 3 sims live, 50+ students, 1 lecturer assigned, assessment data shows improvement, new contributor can onboard from wiki.

4. Technical architecture summary

graph TB subgraph Client Browser[React PWA] Mobile[Flutter wrapper Phase2+] end subgraph Edge CF[Cloudflare Pages] end subgraph Backend API[FastAPI on Render] Cache[Redis cache] end subgraph AI Mistral[Mistral AI] Codestral[Codestral debug] end subgraph Data Supabase[(Supabase PG + Auth)] end Browser --> CF --> API API --> Cache API --> Mistral API --> Codestral API --> Supabase Mobile -.-> API

Key design choices: 3‑tier (React/FastAPI/Supabase), offline PWA, per‑student rate limiting, AI response caching (Redis). All API keys secured on backend.

Database: 15 tables (users, simulation_sessions, quiz_attempts, ai_interactions, etc.) with Row Level Security.

5. Team of 7 – roles & RACI

Tech Lead
Architecture, merges, CI/CD
Frontend Sim
Canvas, algorithms, animation
Frontend UI/UX
React, PWA, responsive
Backend
FastAPI, Supabase, Redis
AI/Prompt Eng
Mistral prompts, caching, fallback
Content & Curriculum
NTA mapping, validation, UAT
QA & DevOps
Playwright, GitHub Actions, monitoring
RACI example: selected activities
ActivityTech LeadFrontend SimBackendAI EngContentQA
Sorting sim CanvasCA/RIICR
AI Tutor APICICA/RIR
Formal assessmentCIIIA/RI
CI/CD pipelineAIRIIR

6. Agile project management & testing

1‑week sprints, asynchronous daily standups (GitHub Discussions), Sunday 30‑min review/retro. Definition of Done includes unit tests + peer review + CI pass. Testing pyramid: 100+ unit tests (Vitest/pytest), integration tests for AI proxy, E2E with Playwright (8 critical journeys).

graph TD subgraph "Testing pyramid" A[E2E – Playwright] --> B[Integration – API + Supabase mocks] B --> C[Unit – Vitest, pytest] end
Risk mitigation: team attrition (handover docs from week1), offline fallback, Mistral free tier with Redis caching, scope lock (one sim before expanding).

7. Cost analysis & sustainability

Monthly infrastructure (MVP): ~$0.10
ServiceTierMonthly cost
Cloudflare PagesFree$0
Render backendFree (sleep)$0
SupabaseFree$0
Upstash RedisFree$0
Mistral AI (200 students)Pay as you go~$0.17

AI semester cost: 20.4M input + 10.2M output tokens → Mistral Small: $1.73 total. GPT‑4o Mini would be $9.18.

pie title "Semester AI cost ($1.73)" "Input tokens (20.4M)" : 0.61 "Output tokens (10.2M)" : 1.12

Scaling: 1,000 students → $25.85/month (Supabase Pro). 10,000 students → $58.50/month. Mistral startup credits ($30k) can be applied for in Phase 3.

The full documentation (30+ sections) includes complete API specs, database ERD, simulation design details, CI/CD workflows, risk register, and onboarding guides. This summary gives you the strategic compass. Now execute.