C-Link
AI-powered campus placement platform linking students, placement cell and recruiters from profile to joining. Local AI advises; humans decide.
- Video of it working
- Problem, approach and result explained
- Public source code
- Team confirmed roles· 5 of 5
- Expert reviewed· 5 reviews
Walkthrough
Overview
The problem
In most colleges, placements run on spreadsheets, WhatsApp groups and manual shortlisting. Students don't know which drives they're eligible for or how to prepare. Recruiters spend hours screening hundreds of candidates, and the placement cell chases updates by hand. Nothing connects profile, screening, interview scheduling, offer and joining in one place. I set out to build one platform for students, the placement cell and recruiters that covers the whole journey. Its rule is that AI advises and humans decide.
The approach
I built C-Link with Next.js, React, TypeScript and PostgreSQL. It has three portals: Student, Placement Cell and Recruiter. Eligibility rules, interview scheduling and clash checks are plain code. The AI is used only where it helps: job-fit scores, preparation plans, drafting screening questions and giving advice on answers. The hardest decision was where AI should not be used. It never selects, rejects or ranks a candidate. For the video interview it reads only the transcript, never the face, and it never judges appearance, accent or eye contact. A silent or unclear answer is flagged for the recruiter to listen to and is not given zero. I also chose to run all AI locally, with Ollama (Gemma) and Whisper, so no student data goes to an outside company. The video interview records one continuous video, stored in PostgreSQL, and only the owning recruiter can play it. It's deleted after 30 days. I chose not to build face or emotion scoring, and not to host recordings publicly.
The result
The platform works end to end. A student applies, takes a text screening test and a recorded video interview, and gets interview slots and offers tracked through to joining. The recruiter reviews the recording, transcript and AI advice on one page, and the placement cell sees progress and analytics. I tested the video interview with a fake camera and real browser recording. The tests covered all questions answered, all skipped, a mix, auto-submit when the time ends, a lost connection, a failed speech or AI service, duplicate submissions, and another recruiter trying to open a video. All of these pass. A 15-minute recording is about 60 MB. Add your screenshots or a clip here: the recruiter's "AI Interviews" page with the player and transcript, and a student recording screen. That's the best way to show it working.
Reflection
I would build a hosted version with proper storage for videos. Right now it runs on a local machine, and videos in the database are fine for one college but would need file storage at a larger scale. Still unfinished: multi-factor login isn't built yet, and I tested transcription quality with a stub, not with many real students' voices. Chrome or Edge on desktop only. The AI marks are advice and need a recruiter's judgment. I learned that the careful choices matter most: what the AI should not do, what happens when a service fails, and how to protect private recordings. Those decisions took more time than the features themselves.
Submitted to CAMPUSLINK: Placement Season Without the Spreadsheet Tandav
- Your idea in one line
C-Link is an AI-powered campus placement platform that connects students, TPOs, and recruiters through intelligent job matching, skill-gap analysis, and placement-readiness tracking using locally hosted AI.
- What will be working by the finale?
1. The full placement pipeline across three role-based dashboards (Student, Placement Cell, Recruiter): TPO roster import (CSV or paste, any column order), recruiter drive posting with AI job-description parsing, rule-based eligibility, and AI matching with an explainable fit score (matched skills, missing skills, written reasoning). 2. Skill-gap identification: AI readiness analysis, preparation plans, and clear reasons when a student isn't eligible. 3. Interview scheduling with clash detection, plus automatic slot allocation across companies, which is the one part still to be built. 4. The offer-to-joining pipeline (request, send, accept or decline, document verification, joined), with in-app notifications for all three roles. 5. Hardening for real use: automated tests for the rules engine and core flow, resume PDF upload, email notifications, and a trial on real college placement data.
- Tech stack and hardware you plan to use
1. Frontend and backend: Next.js 16 (App Router, server actions), React 19, TypeScript, Tailwind CSS 4 2. Database: PostgreSQL with Prisma ORM 3. AI: Gemma (gemma4:e2b) running locally through Ollama, so no student data goes to an external AI service. Used for resume parsing, JD parsing, fit-score explanations and readiness analysis. Eligibility and scheduling rules are plain TypeScript, not AI. 4. Security: JWT session cookies, bcrypt, Zod validation, rate limiting 5. Hardware: Laptop with Ryzen 7, 16 GB RAM and an RTX 5050 (8 GB VRAM) for local AI inference 6. Still needed: A hosted machine to run PostgreSQL and Ollama for the finale demo, and an anonymised placement dataset from our college.
- What have you built or tested so far?
Built and working end to end: 1. All three dashboards: Student, Placement Cell and Recruiter. 2. CSV and paste roster import with flexible column mapping, plus student signup by roll number. 3. AI resume parsing and JD parsing. 4. An explainable fit score, readiness analysis and prep plans. 5. Rule-based eligibility, shown to students with the exact reasons. 6. Special Consideration. 7. The offer-to-joining pipeline, notifications with unread badges, and analytics.
- Pitch deck (PDF)
- Pitch deck (PDF)PDF
- GitHub repository
- github.com/imshahidraza/campuslink
- What's working now?
C-Link is an AI-powered campus-to-corporate placement platform. It brings students, the placement cell (TPO) and recruiters into one system, from profile to joining, and replaces spreadsheets, WhatsApp groups and manual shortlisting. Students: create a profile, upload a resume (PDF), and apply to or withdraw from drives. They get automatic eligibility checks with clear reasons, an AI job-fit score showing matched and missing skills, a readiness check, a preparation plan and a skills roadmap. They can also track their placement journey and get in-app notifications. Recruiters: post drives, shortlist candidates, schedule interviews, send offers and mark joining. The placement cell approves each recruiter first. Two screening rounds are available: Text screening test: multiple-choice questions are scored automatically, written answers get AI comments as advice only, and shortlisting is blocked until the test is done. Recorded video interview: five spoken questions in one continuous recording. Speech is turned into text locally, and AI gives advice on each answer. Recruiters can watch the recording, jump to each answer and read the transcript and marks. Placement cell: manage the student roster, approve recruiters, track drives and offers, view analytics, and schedule interviews with automatic slot allocation and clash handling. Principles and security: AI advises, humans decide. The AI never selects, rejects or ranks anyone and never judges a face or appearance. Passwords are hashed, every action is checked for role and ownership on the server, and recordings are visible only to the owning recruiter and deleted after 30 days. All AI runs locally (Ollama and Whisper), so no student data leaves the machine. Stack: Next.js 16, React 19, TypeScript, Tailwind, PostgreSQL with Prisma. Status: all the features above are built, tested and running locally. Multi-factor login is planned but not yet built.
- GitHub repository
- github.com/imshahidraza/campuslink