Talentro
A full-stack professional networking and skill exchange platform — built like LinkedIn, engineered from scratch
What is it?
Talentro is a full-stack professional networking platform I built from the ground up — designed to help professionals connect, share skills, discover jobs, and grow their network in one cohesive place. Think LinkedIn's core loop: a live feed, connection requests, job listings, and application tracking — but engineered entirely by one developer, on the MERN stack, with serious attention paid to performance, security, and scalability. It's the most architecturally complex project in my portfolio.
What I Did
I was the sole architect, developer, and deployer of Talentro. I designed the entire data model for a social graph from scratch, built the backend API that powers the feed, connections, jobs, and auth systems, and crafted the full React frontend. Critically, I didn't just build features — I identified and fixed real performance bottlenecks (N+1 queries, CLS issues, memory instability) that you only encounter when you're building at social-platform scale.
Tech Stack & Tools
- MERN Stack
- Social Platform
- Full-Stack
- MongoDB Aggregation
- JWT Auth
- Cloudinary
- Next.js
- Performance Engineering
- Personal Project
- Networking Platform
Social Architecture & Data Modeling
Designed MongoDB schemas for a full social graph — users, posts, reactions, nested comments, connections, and job applications — all with referential integrity maintained at the application layer. The connection matrix (pending, accepted, mutual) required careful state modeling to avoid inconsistencies.
Algorithmic Feed System
Built a dynamic feed that surfaces posts from a user's network, ranked algorithmically. Implemented horizontal pagination limits on feed fetching to cap memory usage server-side — a critical stability decision when dealing with large connection graphs and high-volume post data.
Solving the N+1 Query Problem
Discovered severe N+1 query bottlenecks in the search and feed pipelines — where each result triggered additional individual DB calls. Refactored queries using MongoDB aggregation pipelines and selective population to collapse dozens of round-trips into single efficient queries, drastically improving response times.
Professional Networking Engine
Engineered the full connection lifecycle — send invitation, accept/reject, withdraw, remove connection. Built a smart mutual connection suggestion system that traverses second-degree relationships in the social graph to surface relevant people to connect with.
Job Board & Application Tracking
Developed a dedicated jobs portal with algorithmic job recommendations based on user skills, location-based filtering, and keyword search. Users can apply directly on-platform and track the status of all their applications in a personal dashboard.
Media Uploads & Cloudinary Integration
Integrated Cloudinary via Multer middleware for image and video uploads on posts. Handled multipart form data on the server, validated file types and sizes, and stored Cloudinary asset URLs in MongoDB for efficient retrieval.
Frontend Performance — Fixing CLS
Eliminated Cumulative Layout Shifts (CLS) across the feed and profile pages by implementing React skeleton loaders that match the exact dimensions of loaded content — preventing the jarring reflow that kills perceived performance on social feeds. Backed this up with targeted CSS containment optimizations.
Secure Auth with HTTP-Only Cookies
Built authentication using JWTs stored in HTTP-only cookies — not localStorage — making tokens inaccessible to JavaScript and protecting against XSS attacks. Added email verification on signup and a secure password reset flow with short-lived tokens.
- Query Bottleneck Resolved
- N+1
- Collapsed dozens of DB round-trips into single aggregation pipeline queries
- Layout Shift Eliminated
- 0 CLS
- React skeleton loaders matched to exact content dimensions across all feed views
- XSS-Proof Auth
- HTTP-Only
- JWTs in HTTP-only cookies — tokens completely inaccessible to client-side JS
- Social Graph Implemented
- Full
- Feed, connections, mutual suggestions, reactions, nested comments, job board — all from scratch
- Algorithmic post feed with network-aware ranking
- Media posts with image & video upload via Cloudinary
- Custom reaction system on posts
- Nested comment threads
- Full connection lifecycle — invite, accept, withdraw, remove
- Mutual connection suggestions via second-degree graph traversal
- Job board with algorithmic recommendations
- Job filtering by location and required skills
- In-platform job application tracking
- JWT auth via HTTP-only cookies (XSS-proof)
- Email verification on signup
- Secure password reset with short-lived tokens
- React skeleton loaders for zero CLS
- Fully responsive across all devices
Resolved N+1 query bottleneck — collapsed serial DB calls into single MongoDB aggregation pipelines
JWT stored in HTTP-only cookies — not localStorage — blocking XSS token theft entirely
Horizontal feed pagination limits to cap memory usage on large social graphs server-side
Skeleton loaders dimensionally matched to real content — eliminating Cumulative Layout Shift
Second-degree social graph traversal for mutual connection suggestions
Full social platform architecture built solo — feed, graph, jobs, auth, media — no shortcuts
Ready to build something amazing?
Interested in learning more about this project or starting a new full-stack development journey? Let's talk.