Democratizing Personalized Learning with Open Source AI

How do we solve Bloom’s 2 Sigma Problem and scale highly effective 1-to-1 tutoring to every student? While Artificial Intelligence offers a proven solution—significantly boosting student performance and reducing teacher planning hours by 40%—relying on closed-source AI models introduces massive risks.

In this presentation from the FOSSASIA 2026 Summit, I explore why the future of personalized education must be open-source, and how we can architect AI systems that teach students how to think, rather than just giving them the answers.

Transcript

1. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Democratizing Personalized Learning with Open Source AI Kannan Murugapandian March 9, 2026
2. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI
3. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Is AI really effective?
4. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Is AI really effective? Source: Kestin, G. et al. "AI tutoring outperforms in-class active learning." Scientific Reports 15, 17458 (2025). Comparison of mean post-test performance between students taught with the in-class active learning and students taught with the AI tutor. From the Student's Point of View
5. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Is AI really effective? Metric Before AI After AI % Change Weekly Planning Hours 10 6 -40% Satisfaction with Lesson Quality 3.2/5 4.5/5 +40.6% Source: Belloula. IJRES 11(2), 264–280 (2025). Impact of AI Integration on Teachers’ Weekly Planning Hours and Lesson Quality Satisfaction From the Teacher’s Point of View
6. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Why not closed-source AI? • Potential exposure of sensitive data (February 2026) (August 2025) (October 2025)
7. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Why not closed-source AI? • Vendor Lock-In - Dependency Problem
8. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Why not closed-source AI? • Cost Dependency
9. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Fine-tuned Open-Source AI Models
10. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Fine-tuned Open-Source AI Models Task GPT-4 Best Fine-Tuned (7B) Lift over GPT-4 AVERAGE (31 Tasks) 0.661 0.756 +9.5 pts Source: “LoRA Land” — Zhao et al. (2024) · arXiv:2405.00732
11. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Socratic Tutoring • Ask, don't tell: Guide students to discover answers through questions • Why > What: Focus on reasoning, not just correctness • Build autonomy: Students learn how to think, not what to think
12. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Direct Instruction Tutoring Socratic Tutoring
13. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Glass-Box Learning Architecture Student Query Verified Retrieval Socratic Tutor LLM Guardrail Check Final Response + Teacher Insights
14. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Glass-Box Learning Architecture Student Query Verified Retrieval Socratic Tutor LLM Guardrail Check Final Response + Teacher Insights
15. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Glass-Box Learning Architecture Student Query Verified Retrieval Socratic Tutor LLM Guardrail Check Final Response + Teacher Insights
16. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Glass-Box Learning Architecture Student Query Verified Retrieval Socratic Tutor LLM Guardrail Check Final Response + Teacher Insights
17. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Glass-Box Learning Architecture Student Query Verified Retrieval Socratic Tutor LLM Guardrail Check Final Response + Teacher Insights
18. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Glass-Box Learning Architecture Student Query Verified Retrieval Socratic Tutor LLM Guardrail Check Final Response + Teacher Insights
19. Kannan Murugapandian Democratizing Personalized Learning with Open Source AI Thank you for your attention! • Email: mkannan2k9@gmail.com • GitHub: github.com/mkannan2k9 • LinkedIn: linkedin.com/in/mk029