Worm Live Demo URL: https://wormie.app

Worm uses Gemini AI to revolutionize college admissions. Our agentic workflows match students with their perfect college and help universities discover applicants.

Comment

💡 What inspired me

Before I became a full-stack developer, I worked as an international student recruiter for a business school. I witnessed firsthand the massive barrier between prospective students and the admissions office. Students were overwhelmed, stressed, and struggling to navigate fragmented data across dozens of university websites. On the other side, admissions officers were spending valuable time trying to identify high-intent candidates in a sea of generic inquiries.

This friction made me realize that the higher education discovery process was fundamentally broken. I wanted to build a solution that worked for both parties: an intelligent platform that acts as a personalized guide for the student, while seamlessly serving as a proactive lead-generation tool for the university. When I saw the capabilities of Google Gemini, I knew it was the perfect engine to finally bridge this gap.

💻 What it Does

Wormie delivers a complete three-pillar higher education ecosystem:

  • For Students (The AI Counselor & Marketplace):

    • Wormie AI: An autonomous counselor powered by Gemini 2.0 Flash that answers complex admissions questions, evaluates GPA/SAT chances against official benchmarks, and autonomously bookmarks colleges.
    • Advisor Marketplace: Students can browse verified admissions advisors, view real-time availability calendars, and book 1-on-1 video consultations with built-in WebRTC video calling.
    • Social Community: Dedicated college discussion hubs and real-time direct messaging to connect with peers and current university students.

  • For University Recruiters (B2B Recruitment CRM & Lead Pipeline):

    • Recruiter Portals: Verified admissions staff can access dedicated applicant dashboards showing matching candidate pipelines.
    • Autonomous Lead Routing: When a student expresses interest during a counseling session, Gemini executes an autonomous tool call that seamlessly routes their verified profile into the institution's CRM as an active recruit lead.

  • For Administrators & Advertisers (B2B Ad Monetization & Telemetry Engine):

    • Ad Management Dashboard: A high-performance advertising and operations dashboard where administrators can launch targeted campus spotlight campaigns.
    • Flexible Monetization Models: Supports Flat Rate, Cost-Per-Click (CPC), and Cost-Per-Mille (CPM), complete with real-time impression/click tracking and ROI metrics.

🛠️ How I built my project

Wormie is built from the ground up to integrate AI into user interactions, database state management, and business telemetry:

  • Gemini 2.0 Managed Agent with Native Function Calling: Powered by the official google-genai SDK, Wormie uses Gemini 2.0's function calling to execute multi-step database actions autonomously. We defined typed Python tool functions with descriptive docstrings (get_college_details, search_colleges, calculate_admission_chances, bookmark_college, submit_lead) that Gemini inspects, decides to invoke, and evaluates in a multi-step loop.
  • Full-Stack Architecture:
    • Frontend: Built with React, featuring a sleek, responsive design system, real-time message streaming, and an integrated WebRTC video call room.
    • Backend: Engineered with Django & Django REST Framework, running on PostgreSQL with optimized query filters across 7,000+ U.S. colleges from the Department of Education (IPEDS) dataset.
    • Telemetry & Analytics: Built-in telemetry tracking engine that logs AI latency, token counts, error rates, and ad engagement events (impressions and click-through rates).
  • Cloud Infrastructure: Fully deployed to production with containerized backend services on Google Cloud Run, managed database on Cloud SQL (PostgreSQL), and static client hosting on Firebase Hosting.

🧗 The challenges I faced

The biggest technical hurdle was transitioning from a basic completion model to an autonomous agent with database agency. Because Wormie triggers actual database writes (like booking a recruiter lead or saving a bookmark), I engineered strongly-typed tool closures with strict validation to ensure the model only executes actions when the user demonstrates clear, explicit intent, preventing hallucinations from cluttering the recruiter's CRM.

Additionally, streaming the Gemini response to the React frontend while simultaneously executing backend tool calls required careful asynchronous pipeline management in Django to ensure the UI felt completely seamless and instantaneous.

🧠 What I learned

This project fundamentally changed how I view LLMs. I learned that generative AI is exponentially more powerful when it transitions from being a passive "chatbot" to an active "agent" with agency over a database. By mastering RAG pipelines and native tool calling with the Gemini API, I learned how to turn unstructured natural language into structured, high-value business operations.

🔮 What's next for Wormie

  • Mobile Native App: Releasing native iOS and Android apps using React Native.
  • Agentic Scholarship Discovery: Expanding Gemini tool functions to query federal and private financial aid datasets.
  • CRM Integrations: Building native webhooks to sync leads directly with standard university CRMs like Slate and Salesforce for Education.

Built With

Share this project:

Updates

Submission history