Inspiration

• Existing AI tutors generate answers but rarely measure real learning or provide evidence that learning occurred.
• We wanted to build an adaptive learning platform that understands each student's knowledge and produces cryptographically verifiable learning progress.

What it does

• Assesses student knowledge through adaptive diagnostic tests.
• Creates a personalized Digital Twin representing mastery, confidence, XP, and learning history.
• Uses a Knowledge Graph to understand prerequisites and recommend the next best concept.
• Generates personalized learning roadmaps, AI tutoring sessions, and adaptive practice questions.
• Continuously updates the learner model after every interaction.
• Uses Finality to verify important learning events with hashing, signatures, replay protection, and audit trails, making progress tamper-evident.

How we built it

• Built as a TypeScript monorepo using Turborepo and pnpm.
• Frontend powered by Next.js, React, and Tailwind CSS.
• Backend data layer built with PostgreSQL and Drizzle ORM.

• Modular architecture with independent packages:

  • Shared
  • Database
  • Knowledge Graph
  • Diagnostic
  • Digital Twin
  • Roadmap
  • Tutor
  • Practice
  • Finality
  • Finality Adapter

• Designed every package to be reusable, testable, and independently scalable.

Challenges we ran into

• Designing and integrating a large modular architecture within a short hackathon timeline.
• Synchronizing state across Diagnostic, Digital Twin, Knowledge Graph, Roadmap, Tutor, Practice, and Finality.
• Creating a reusable verification pipeline that secures learning events without disrupting the learning experience.
• Balancing AI flexibility with deterministic learning logic.

Accomplishments that we're proud of

• Built an end-to-end adaptive learning workflow from assessment to verified learning.
• Designed a scalable architecture with multiple reusable packages.
• Integrated cryptographic verification through Finality, making Sakshion more than a traditional AI tutor.
• Created a strong technical foundation that can evolve into a production-ready personalized learning platform.

What we learned

• Personalized education works best when AI is guided by structured knowledge rather than isolated prompts.
• Combining Knowledge Graphs, Digital Twins, and adaptive practice creates a continuous learning feedback loop.
• Verifiable AI interactions increase trust and transparency in educational systems.
• A modular architecture greatly simplifies development, testing, and future expansion.

What's next for Sakshion

• Complete frontend integration and interactive dashboards.
• Deploy the platform with authentication and cloud infrastructure.
• Expand support for additional subjects and learning content.
• Add real-time voice and multimodal AI tutoring.
• Build teacher, classroom, and institutional analytics dashboards.
• Enable verified learning records and achievements that can be securely shared with schools, employers, and certification providers.

Built With

• monorepo
• neon
• nextjs
• openai