Inspiration
Students often know what they have to study, but not what they should study next.
We built RUET Pulse to solve that problem. Instead of being another AI chatbot that simply answers questions from uploaded documents, RUET Pulse builds an academic picture of the student: what they know, where they are weak, what is important in their courses, and what deserves their limited study time next.
The idea came from our own experience as university students. With multiple courses, lecture materials, exams, and limited time, the difficult part is not finding information — it is deciding where to spend the next 30 minutes.
What it does
RUET Pulse turns a student's course materials into an adaptive academic decision system.
Students can:
- Add their courses, exam dates, and daily study budget.
- Upload lecture notes, slides, and course PDFs.
- Use Gemini to extract important concepts from their materials.
- Take an AI-generated diagnostic assessment.
- See actual performance at the concept level.
- Identify weak and developing areas.
- Learn weak concepts through an adaptive AI study session.
- Practice and reassess their understanding.
- Receive a Next Best Action based on mastery, course importance, exam proximity, mistakes, and available study time.
The goal is simple:
Don't just tell students what is in their notes. Tell them what they should do next.
What makes it different
Tools such as document-based AI assistants are excellent at helping students explore and understand information.
RUET Pulse focuses on a different question:
"Given everything I have to study and the time I have available, what should I do next?"
The application separates AI reasoning from academic decision logic. Gemini handles tasks such as concept extraction, question generation, explanations, and tutoring, while deterministic application logic calculates mastery, priority, exam proximity, and the student's next best action.
This makes RUET Pulse less like a document chatbot and more like an AI academic decision engine.
How we built it
We built the application as a web-based academic assistant with a structured state model for courses, materials, concepts, assessments, mastery, and study sessions.
Course materials are analyzed into compact structured knowledge rather than repeatedly sending entire documents to the model. Diagnostic questions are linked to individual concepts, allowing the system to determine where a student is actually struggling.
The system then combines:
- Concept mastery
- Course importance
- Exam proximity
- Recent mistakes
- Prerequisites
- Daily study budget
to calculate a prioritized Next Best Action.
Gemini is used where generative intelligence is most valuable: understanding academic material, generating assessments, explaining concepts, and adapting teaching to the student's performance.
Challenges we faced
One of our biggest challenges was making the AI pipeline reliable enough for real academic documents.
During development we encountered API quota limits, transient model availability errors, document upload failures, response-format mismatches, and state synchronization issues between material analysis and diagnostic generation.
We solved these by implementing sequential material processing, controlled retry and backoff strategies, strict JSON API contracts, document-level analysis states, and a centralized academic state model.
Another challenge was ensuring that AI-generated results actually affected the student's academic state instead of becoming isolated chatbot responses.
That led us to design RUET Pulse around a continuous learning loop:
Analyze → Assess → Measure → Teach → Practice → Reassess → Prioritize
What we learned
The most important lesson was that adding AI to an application is not enough.
A useful AI product needs a clear decision loop around the model.
Instead of asking Gemini to decide everything, we gave the application responsibility for measurable academic state and used Gemini where contextual reasoning and generation are strongest.
The result is an AI system designed not just to answer students, but to help them make better academic decisions.
Built With
- academic
- adaptive
- ai
- assistant
- edtech
- education
- learning
- personalized
- productivity
- student
- tutor
- typescript