Samur.ai

Samur.ai Logo
Landing
Review
Review (Timeline)
Review (Edit)
Export

Inspiration We noticed that every semester students waste hours manually copying deadlines from PDF syllabi into their calendars. We wanted to solve the problem of missing assignments due to clutter and disorganization by automating the tedious part of semester planning.

What it does Samur.ai takes a course syllabus via text paste or PDF upload and uses AI to extract key dates like assignments and exams. It organizes these into a reviewable list with confidence scores and allows users to export them directly to Google Calendar, Outlook, or as a standard ICS file.

How we built it We built the frontend with Next.js and React for a responsive interface. The core logic uses the Google Gemini Flash 2.5 API to parse unstructured syllabus text and identify temporal patterns. We used PDF.js for file parsing and Tailwind CSS to implement the custom visual theme.

Challenges we ran into Handling recurring events like homework due every Friday was difficult to structure logic for. We also struggled with extracting text cleanly from complex PDF layouts without losing the context of the dates found in tables.

Accomplishments that we're proud of We are proud of the confidence scoring system that flags uncertain dates for user review. We also successfully implemented a clean drag and drop PDF uploader that feeds directly into the generative model for high accuracy extraction.

What we learned We learned that prompt engineering is critical when asking LLMs to output strict JSON formats. We also discovered that providing the model with raw PDF data is often more effective than pre-processing the text ourselves.

What's next for Samur.ai We plan to integrate directly with LMS platforms like Canvas or Blackboard. We also want to add smart study scheduling that automatically blocks out preparation time before major exams.