Vertex Platform

Inspiration As engineering students, we constantly faced the "Black Hole" of hiring. We sent out dozens of CVs, only to be "ghosted" by automated ATS systems that filter based on keywords, not actual skills.

We ran a survey of over 50 students and recruiters and found a shocking disconnect: 85% of students feel their practical projects are ignored, while 90% of recruiters spend less than 30 seconds on a CV because they don't trust the self-reported skills.

We realized the recruitment system is broken. It rewards people who are good at writing CVs, not people who are good at writing code. We built Vertex to change that—replacing "Who you know" with "What you can actually do."

What it does Vertex is an AI-powered bridge between academic potential and industry needs. It is a two-sided platform:

For Students (The Verified Profile): Instead of a static PDF, students connect their GitHub. Vertex uses Gemini 3 Pro to analyze their actual code repositories, verifying their skills (e.g., "This student actually knows FastAPI because they built a working auth system"). It also parses their PDF CVs using multimodal AI to understand visual layouts.

For Recruiters (Natural Search): Companies stop searching for keywords. Instead, they ask questions like: "Find me a junior developer who knows React and has built a real-time chat app." Vertex returns a ranked list of verified candidates.

OpenMatch: A smart matchmaking system that helps students find teammates for graduation projects and hackathons based on complementary skill sets.

How we built it We treated this hackathon as a professional product sprint, utilizing a unique "Antigravity" Workflow to allow four developers to code simultaneously without conflicts:

The Brain (AI): We used Google Gemini 3 Pro via the Python SDK. We leveraged its large context window to feed it entire GitHub README.md files and code snippets for deep analysis.

The Backend: Built with FastAPI (Python) for high-performance async processing. We used PostgreSQL on Railway for our database, managing complex relationships between Users, Skills, and Projects.

The Frontend: A responsive UI built with React, Vite, and Tailwind CSS, deployed on Vercel.

Architecture: We adopted a "Modular Monolith" structure, separating the codebase into distinct domains (Auth, Students, Company, AI) so every team member had full ownership of their layer.

Challenges we ran into The "Local vs. Cloud" Database Trap: We struggled initially to connect our local DBeaver clients to the Railway database while keeping our backend secure. We learned the hard way about the difference between Railway's "Public Proxy" (for us) and "Private Network" (for the app).

AI Hallucinations: Early on, the AI would "invent" skills for students. We fixed this by implementing strict Pydantic schemas and lowering the temperature for the Gemini API, forcing it to cite specific files as evidence for every skill it verified.

Deployment Integration: Connecting a Vercel Frontend to a Railway Backend required careful CORS configuration and environment variable management, which was a major learning curve for the team.

Accomplishments that we're proud of True "Vibe Coding": We built a production-ready Full Stack application in days by leveraging AI for boilerplate while we focused on the complex logic.

Multimodal Parsing: We didn't just extract text; we built a pipeline that understands the structure of a developer's portfolio.

Live Deployment: Vertex isn't just running on localhost; it is fully deployed and live, with a persistent database and real-time API.

What we learned Prompt Engineering is Logic: We learned that getting good results from Gemini isn't about magic words; it's about providing structured context and clear constraints.

Database Design Matters: Spending time on our ER Diagram (Entity Relationship) on Day 1 saved us hours of debugging later.

Team Velocity: We learned that a good workflow (Role A/B/C/D) is just as important as good code.

What's next for Vertex Platform Video Interview Analysis: Using Gemini 1.5 Pro's video capabilities to analyze mock interviews and give students feedback on their soft skills.

LMS Integration: Connecting directly with University portals to pull verified academic grades.

Global Hackathon Mode: Opening the "OpenMatch" feature to cross-university hackathons to foster global collaboration.