Tagline
Guaranteed Graduation Path, Instantly.
🌟 Inspiration
Academic advising is fundamentally broken. University course catalogs are cryptic, unstructured PDFs, forcing students to manually juggle complex prerequisite chains, credit minimums, and general education outcomes. This error-prone process leads to scheduling mistakes, missed requirements, and, ultimately, wasted tuition money and delayed graduation. We were inspired to solve the core technical challenge: making the official degree rules instantly legible and actionable.
💻 What it does
FourSight is a proof-of-concept MVP that demonstrates the future of academic advising. It takes a raw, unstructured university degree requirements document (like a PDF or catalog text) and transforms it into a guaranteed, prerequisite-compliant 4-year course schedule.
The user provides their:
- Major requirements file.
- Detailed prerequisite list.
- Desired semester credit load.
FourSight then outputs an optimized, semester-by-semester plan that meets every dependency and hits the final credit goal.
🛠️ How we built it
The pipeline is built on the reasoning and parsing power of Google Gemini.
- Catalog Parsing: We use Gemini's LLM capabilities, guided by highly specific prompt instructions and custom regex logic, to reliably extract structured data (course titles, credit hours, and credit totals) from the messy, unstructured text of the academic catalog.
🛑 Challenges we ran into
The primary challenge was handling the inherent ambiguity and inconsistency of real-world academic documents. Specifically:
- Complex Grouping: Extracting requirements like "One course from MTH 215, 243, 244 or STA 307" and representing it as a single requirement in our data structure.
- Prerequisite Specificity: Fine-tuning the regex (e.g., modifying
r"Pre:\s*(.*?)(?:\.|$)"to ignore periods in GPA or version numbers) to avoid prematurely cutting off valid prerequisite text. - Credit Balancing: Generating an output that respects prerequisite order while maintaining a near-perfectly consistent credit load (e.g., 15 credits) across all 8 semesters.
✨ Accomplishments that we're proud of
Our biggest accomplishment is proving the core concept: that an LLM can successfully and reliably extract the necessary structure from a real-world, complex university PDF to drive a deterministic scheduling algorithm. We successfully moved beyond simple text summarization to generating actionable, structured data that solves the prerequisite-chaining problem—the single hardest constraint in academic advising.
🎓 What we learned
We learned that LLMs like Gemini are transformative tools for structured data extraction from non-standardized sources, a task previously reserved for brittle, hand-coded scripts. This unlocks massive potential for the modernization of university administration and student services. We shifted the problem from "Can we read the document?" to "Can we solve the scheduling logic?"—a far more exciting and scalable problem.
⏭️ What's next for FourSight
To evolve from an MVP to a fully functional tool, we plan to:
- Dynamic Elective Integration: Build an API to fetch real course titles that fulfill General Education outcomes (e.g., A1, B2) and integrate user interests to populate elective slots.
- Semester Availability Constraint: Integrate data on which courses are only offered in Fall or Spring to make the plan truly realistic.
- Direct Registration Interface: Connect the finalized 4-year plan to a simulated registration system, turning the plan into a one-click enrollment solution.
Log in or sign up for Devpost to join the conversation.