Inspiration Traditional studying is often a slog—static textbooks, endless highlighting, and the overwhelming anxiety of "where do I start?". We wanted to turn that friction into agency. What if preparing for a high-stakes competitive exam felt like a tactical mission? What if understanding a complex concept felt like unlocking a secret level? We were inspired by the gamification of productivity and the realization that AI should be more than a passive search engine—it should be a "Study Commander" that actively coaches and challenges you. What it does StudySmartAI (EXAM COMMAND) is an all-in-one multimodal learning hub. Users can paste dense lecture notes, upload a photo of a handwritten syllabus, or simply search for a topic. The system instantly synthesizes this data into an "Elite Mission Briefing": Executive Intel: High-level summaries for rapid priming. Concept Clarity (ELI5): Complex topics broken down into simple words and analogies. Key Terms: Interactive flashcards for active recall. Combat Test: Practice questions that evaluate performance and provide targeted fixes for weaknesses. The Study Plan: A personalized, day-by-day roadmap leading to the exam. Boss Battle: A cinematic, turn-based text adventure where users fight "Curriculum Bosses" by proving their knowledge in real-time. How we built it The stack was engineered for speed, logic, and cinematic flair: Frontend: React 19 and TypeScript, using Tailwind CSS for a high-contrast, terminal-inspired aesthetic. AI Intelligence: Powered by Gemini 3 Flash. We leveraged its massive context window and native multimodal capabilities to process both complex text and image data (OCR) in a single pass. System Architecture: We implemented a "Persona-Switching" engine. The AI operates in Study Coach Mode for generating materials and Cinematic Director Mode for running the game logic, ensuring a consistent tone for each phase. Logic: Custom state machines manage the "Neural Link" (processing state) to provide a smooth transition from raw input to the interactive hub. Challenges we ran into Structured Data Extraction: Getting an LLM to consistently return complex, nested JSON (combining study plans, test questions, and game narrative) required rigorous schema enforcement and prompt engineering to avoid parsing failures during "Neural Compilation." Cinematic Pacing: Balancing the technical accuracy of exam questions with the narrative flow of a "Boss Battle" was tricky. We had to fine-tune the system instructions so the boss's dialogue reacted dynamically to whether the user’s answer was right or wrong. OCR Reliability: Extracting structured meaning from varying quality of note images required precise multimodal prompting to ensure the generated "Combat Test" actually matched the uploaded syllabus. Accomplishments that we're proud of The "Game-Study" Loop: We successfully built a tool where the user wants to answer the next question, not just to learn, but to see the next cinematic scene and deal "damage" to the boss. Immersive UI: The "Command Terminal" aesthetic—complete with camera-shake effects on boss attacks and glowing indigo accents—creates an atmosphere of focus and intensity that standard dashboards lacks. One-Click Intelligence: The ability to go from a blank screen to a 7-day study plan and a full suite of materials in under 5 seconds. What we learned We discovered that UI as Narrative significantly increases engagement. By framing study tasks as "Missions" and "Intel," the psychological barrier to starting is lowered. Technically, we mastered the use of Gemini's responseSchema to handle highly complex data structures, which is critical for building stable AI-driven applications. What's next for StudySmartAI Live Audio Interrogations: Integrating Gemini's Live API for voice-based "oral exams" where the AI grills you on concepts while you're on the move. Co-op Missions: Real-time study sessions where multiple students can join a lobby to take down a "Final Boss" (like a mock SAT or GRE) together. The Knowledge Vault: Long-term memory storage where the AI tracks your performance across different subjects over months to identify multi-disciplinary weaknesses.
Built With
- firebase-(auth-&-firestore)
- font-awesome-6
- google-fonts-(space-grotesk)
- google-gemini-api-(gemini-3-flash-preview)
- react-19
- tailwind-css
- typescript
Log in or sign up for Devpost to join the conversation.