Kognos

An elite, multimodal AI mentor for high-stakes 2026 exams like JAMB, WAEC, GRE, and IELTS. Uses the Socratic method and real-time voice/vision to guide students toward conceptual mastery.

Comment

Inspiration

Education often feels like a fragmented experience. Students jump between static PDFs, complex datasets, and frantic web searches, losing the "flow" of deep learning. We wanted to build a unified, intelligent cockpit for students, specifically those tackling high-stakes exams like JAMB, WAEC, and the GRE. We were inspired by the idea of an AI that doesn't just give answers, but acts as a multimodal study partner that understands the logic behind the curriculum.

What it does

Kognos is a multimodal AI study platform designed for the "Marathon Mission" of academic preparation. It serves as a centralized hub where students can:

  1. Analyze Multimodal Content: Native analysis of PDF study aids and complex Data Sheets (ACCA/GMAT datasets).

  2. Study Web Integration: Capture diagrams and external data directly from the web into a unified workspace.

  3. Live Logic Tracking: The platform organizes study milestones into "Logic" paths (e.g., Calculus Applications, General Knowledge) rather than just folders.

  4. Real-Time Exam Alerts: Keeps students on track with a "Live Alerts" dashboard for registration deadlines and curriculum changes.

How we built it

We built Kognos to be fast, dark-mode optimized, and deeply integrated.

Frontend: A sleek, high-contrast React-based interface designed to minimize eye strain during long "marathon" study sessions.

Intelligence Layer: We leveraged Gemini’s multimodal capabilities to process both text and visual diagrams within textbooks simultaneously.

Socratic Engine: For the "Data Sheets" feature, we implemented a Socratic breakdown method that guides users through spreadsheet data rather than just summarizing it.

State Management: A custom sidebar tracking system to manage "Core Milestones" and "Live Alerts" ensuring the user always knows their progress in the 2026 academic cycle.

Challenges we ran into

One of the biggest hurdles was multimodal context switching. Getting the AI to maintain the "Logic" of a Calculus problem while referencing a specific diagram from an uploaded PDF required fine-tuning our prompt orchestration. Additionally, designing a UI that provides a lot of information (deadlines, milestones, tools) without feeling cluttered was a constant balancing act.

Accomplishments that we're proud of

  1. Seamless PDF-to-Logic Pipeline: We successfully created a system where a user can upload a massive WAEC/GRE book and have it broken down into "Logic" steps almost instantly.

  2. The "Study Web" Feature: This tool bridges the gap between static local files and the vast information on the internet, allowing for real-time diagram capture.

  3. User-Centric Design: Creating a professional, "mission-control" aesthetic that makes the user feel like an elite candidate preparing for a mission, rather than just a student doing homework.

What we learned

We learned that context is king in education. It’s not enough for an AI to be smart; it has to be aware of the user's specific timeline (like the Feb 28th JAMB deadline). We also discovered that "Socratic" AI interaction, where the AI asks questions back to the student leads to much higher retention rates than simple chat-and-response.

What's next for Kognos

The journey for Kognos is just beginning. Our roadmap includes:

  1. Collaborative "Live Rooms": Turning the "Live Room" button into a fully functional peer-to-peer study space with shared AI whiteboards.

  2. Predictive Analytics: Using the "Core Milestones" data to predict student performance and suggest areas that need more focus before exam day.

  3. Mobile Live Mode: Bringing Kognos to mobile with camera-sharing so students can point their phone at a physical textbook and bring it into their digital logic path.

GEMINI INTEGRATION

Kognos is powered by Gemini 3 Flash, leveraging its frontier reasoning and multimodal capabilities to transform the study experience. Two specific features of the Gemini 3 family are central to the application:

  1. Dynamic Multimodal Reasoning (thinking_level) Kognos utilizes Gemini 3's new thinking_level parameter to modulate cognitive effort. For the Data Sheets feature, we set the thinking level to "high", enabling the model to perform deep, internal reasoning over complex ACCA/GMAT datasets before providing a Socratic breakdown. Conversely, for Live Alerts and quick registration queries, we use the "minimal" level to ensure lightning-fast latency (3x faster than previous versions), keeping the student’s flow uninterrupted.

  2. Native Document Understanding (Layout-Aware Parsing) Unlike traditional OCR, Gemini 3 Flash treats PDFs as structured multimodal documents. In Kognos, this is central to our Logic Paths: the model understands the spatial relationship between text, complex diagrams, and tables within a textbook. This allows Kognos to explain a Calculus diagram's visual logic while simultaneously referencing the surrounding text, providing a cohesive study partner that "sees" the page exactly as the student does.

By combining this Pro-grade intelligence with Flash-level speed and context caching, Kognos provides a cost-effective, high-performance cockpit for high-stakes exam preparation.

Built With

Share this project:

Updates