Scorpio
Transforming Physics Education through AI-Driven Socratic Tutoring.
Scorpio is a research-driven educational platform engineered to transform physics instruction. By integrating a novel 4-layer AI constraint architecture with a high-performance, space-themed interface, Scorpio bridges the gap between traditional Learning Management Systems (LMS) and the dynamic cognitive requirements of physics problem-solving.
Empirical Benchmark Results
Technical Architecture
Scorpio’s architecture is built for concurrency, type safety, and real-time synchronization.
The 4-Layer Constraint System
At the heart of Scorpio’s AI tutoring capabilities is a proprietary constraint architecture designed to ensure pedagogical validity and eliminate direct-answer harvesting without expensive weight fine-tuning. The system enforces strict adherence through four inference-time layers:
Restricts model responses exclusively to physics principles and blocks adversarial prompt injection or subject drift.
Enforces the Socratic method; direct numerical and symbolic solutions are prohibited in favor of targeted conceptual scaffolding.
Mandates proper LaTeX mathematical formatting () and strict SI unit adherence across all derivations.
Synchronizes all constraint layers with the active problem state in Firestore to guide students within their Zone of Proximal Development.
Ablation Progression Across Constraint Tiers
Data & Synchronization
- Real-Time State: Powered by Cloud Firestore with optimistic UI updates for zero-latency interaction.
- Asset Management: Client-side processed submissions (PDFs/Images) stored via optimized base64 in Firebase Storage.
Core Capabilities
Research-Grade AI Tutoring
The Scorpio AI Tutor is a context-aware pedagogical agent, not a generic chatbot.
- Context Retention: Maintains deep awareness of specific assignments and problem states.
- Adaptive Guidance: Dynamically scales hint complexity from concise high school prompts (~477 characters) to advanced college analytical mechanics (~3,316 characters).
- Academic Integrity: Hardened against jailbreak attempts and direct-answer harvesting.
Advanced Mathematical Rendering
Precision is non-negotiable in physics. Our custom rendering engine includes:
- KaTeX Integration: Blazing fast client-side LaTeX rendering.
- Visual Math Builder: An intuitive UI for constructing complex equations without raw LaTeX knowledge.
Immersive User Experience
A “Space-Themed” aesthetic designed to reduce cognitive load and boost engagement.
- Parallax Depth: Multi-layered backgrounds powered by Framer Motion.
- Glassmorphism: Context-aware backdrop-blur interfaces using Radix UI primitives.
Technology Stack
| Layer | Technology | Purpose |
|---|---|---|
| Frontend | Next.js 15 | Server Components, Streaming, and Routing |
| Language | TypeScript 5.7 | Strict type safety and developer ergonomics |
| Styling | Tailwind CSS | Utility-first styling with perceptually uniform colors |
| UI Library | Shadcn UI | Accessible, headless component primitives |
| Motion | Framer Motion | Physics-based animations and gesture handling |
| Backend | Firebase | Auth, Firestore (NoSQL), Functions, Storage |
| AI Model | Gemini 2.5 Flash | Multimodal reasoning and constraint adherence |
| Math | KaTeX | Fast, accessible equation rendering |