# Sumit Yadav — Full Personal & Technical Profile

> Sumit Yadav: Computer Engineering student and full-stack MERN developer in India. Real-time and AI web apps. Open to web development internships.

## Identity & Contact
- **Name**: Sumit Yadav
- **Email**: sumit.yadav.0287@gmail.com
- **Location**: India
- **Status**: Open to work (Seeking web development internship)
- **Portfolio URL**: https://www.sumity88.in
- **Resume (PDF)**: https://www.sumity88.in/resume.pdf
- **GitHub**: https://github.com/Sumit-y88
- **LinkedIn**: https://www.linkedin.com/in/sumit-yadav-ab538b321/
- **Codolio**: https://codolio.com/profile/Sumit_y88
- **Instagram**: https://www.instagram.com/sumit_y88

## Executive Manifesto
I'm a third-year Computer Engineering student and full-stack developer building practical web applications with MERN and Next.js. I enjoy solving engineering challenges beyond the UI, from authentication and real-time collaboration to API design and deployment. I'm looking for a web development internship where I can contribute to real products, learn from experienced engineers, and ship meaningful features.

## Academic Timeline
- **Now**: Computer Engineering, Semester 5 — CGPA 9.81 / 10 (Affiliated with University of Mumbai · Top academic standing)
- **Community**: Active Technical Member — Member (Student Chapter · Collaborative technical events and developer initiatives)
- **HSC Science**: Class XII (Science) — 84.67% (Focus on Physics, Chemistry, and Mathematics)
- **SSC**: Class X (Maharashtra Board) — 92.40% (Maharashtra State Board of Secondary Education)

## Frequently Asked Questions (Quick Facts)
- **Q: who is sumit yadav?**
  **A:** A third-year Computer Engineering student and full-stack developer based in India.
- **Q: what does he build?**
  **A:** Real-time collaborative and AI-powered web applications using MERN and Next.js.
- **Q: what is his stack?**
  **A:** React, Next.js, Node.js, Express, MongoDB, Socket.IO, Yjs, Redis, Docker, and Java.
- **Q: is he available?**
  **A:** Yes, actively seeking a web development internship.
- **Q: where is he based?**
  **A:** Mumbai, India (IST / UTC+5:30).
- **Q: how to contact him?**
  **A:** Directly via email at sumit.yadav.0287@gmail.com.
- **Q: where is his resume?**
  **A:** Available as a verified PDF: /resume.pdf.

## Technical Capabilities
### Frontend
*Modern interfaces & responsive design*
- **Technologies**: HTML, CSS, JavaScript, React.js, Next.js, Vite, Tailwind CSS (learning)
- **Summary**: Building fast, responsive user interfaces with semantic HTML, modern CSS architectures, React, Next.js App Router, and Vite tooling.

### Backend & Data
*Scalable services & clean schemas*
- **Technologies**: Node.js, Express.js, REST APIs, MongoDB, JWT Auth
- **Summary**: Designing secure REST APIs with Node.js and Express, implementing JWT auth flows with refresh tokens, and managing structured documents in MongoDB.

### Real-time & Deployment
*Synchronization & infrastructure*
- **Technologies**: Socket.IO, Yjs, Redis, Docker, Nginx, Git, GitHub Actions, Vercel, Render
- **Summary**: Architecting real-time CRDT collaboration with Socket.IO and Yjs, Redis rate-limiting, containerizing environments with Docker and Nginx, and automating CI/CD.

### AI & Problem Solving
*Algorithms & LLM integration*
- **Technologies**: Google Gemini API, Java, Data Structures & Algorithms (200+ Solved)
- **Summary**: Integrating generative AI via Google Gemini API for automated analysis and practicing rigorous algorithmic problem solving with 200+ problems in Java.
- **Reference**: [View Codolio Profile](https://codolio.com/profile/Sumit_y88)

## Engineering Process
- **Step 01 — Understand**: Pin down what needs to exist and who uses it.
- **Step 02 — Build**: Get a working version early, then harden it: auth, limits, logging.
- **Step 03 — Ship**: Deploy it, link it, fix what breaks.

## Selected Work & Case Studies
### MergeCanvas (Real-time Web Application)
*Real-time collaborative whiteboard with live rooms, cursors, and conflict-free CRDT sync.*
- **Stack**: Next.js 16, React 19, TypeScript, Tailwind CSS, Pusher, Yjs CRDT, Node.js, Canvas API
- **Live Demo**: https://merge-canvas.vercel.app
- **Source Code**: https://github.com/Sumit-y88/Merge-canvas

#### Architecture & Deep Dive
MergeCanvas is a collaborative whiteboard application where teams can draw, sketch, and brainstorm together in real time. Built as a unified full-stack Next.js app with live cursors, persistent rooms, and conflict-free state synchronization.
> "Draw, sketch, and brainstorm together — live cursors and deterministic state convergence."

**System Architecture:**
- **Client**: Next.js 16 App Router, React 19, HTML Canvas API, Tailwind CSS
- **REST API**: Next.js Route Handlers for room initialization and persistence
- **Realtime**: Bidirectional WebSocket event streaming with live cursor broadcast and presence
- **Database**: MongoDB for rooms, user accounts, and persistent board state snapshots
- **Authentication**: JWT authentication with verified session handling

**Production Readiness & Engineering Hardening:**
- Real-time multi-user cursor tracking with debounced position broadcasting
- Persistent room state serialization allowing users to leave and resume whiteboard sessions
- Request and canvas payload boundaries enforced on API handlers
- Responsive viewport scaling for desktop, tablet, and mobile collaboration
- Deployed to Vercel with automated continuous delivery

**Key Architectural Decisions:**
- **Deterministic Conflict Resolution**: Ensures concurrent strokes and shape transformations merge cleanly across clients without race conditions or state loss.
- **Unified Next.js Full-Stack Architecture**: Combines canvas rendering, server API routes, and real-time coordination inside a single maintainable repository ready for edge deployment.

### InfraChat (MERN + Socket.IO Platform)
*Role-based internal messaging platform with room management and a 6-stage command pipeline.*
- **Stack**: React 19, Node.js, Express, Socket.IO, MongoDB, Tailwind CSS, JWT Auth, RBAC
- **Live Demo**: https://infra-chat-six.vercel.app
- **Source Code**: https://github.com/Sumit-y88/Infra-Chat

#### Architecture & Deep Dive
InfraChat (The Internal Network) is a real-time internal messaging platform engineered for distributed developer and engineering teams. Powered by the MERN stack and Socket.io, it provides instant room-based and direct messaging, typing presence, audit logging, and an extensible slash-command execution engine.
> "Secure internal communication powered by 3-tier RBAC and a 6-stage command execution pipeline."

**System Architecture:**
- **Client**: React 19 single-page application styled with Tailwind CSS
- **REST API**: Express REST API for authentication, user management, and room configurations
- **Realtime**: Socket.io 4.8 duplex gateway managing room broadcasts, presence, and live delivery
- **Database**: MongoDB 7.0+ storing message histories, room permissions, user roles, and audit trails
- **Authentication**: 3-tier Role-Based Access Control (SuperAdmin, Moderator, Member) with signed JWT tokens

**Production Readiness & Engineering Hardening:**
- 6-stage pipeline (Parse → Lookup → Permission → Context → Cooldown → Execute) for slash commands
- Granular role-based moderation enabling kick, ban, mute, and room creation actions
- Online presence tracking and typing indicator events throttled across active rooms
- Media upload handling with size verification up to 10MB
- Structured audit logs capturing administrative actions across public and private channels

**Key Architectural Decisions:**
- **6-Stage Command Pipeline**: Intercepts slash commands through an isolated pipeline enforcing cooldowns and role validations before server execution to prevent socket spam and privilege escalation.
- **Hierarchical Role-Based Security**: Strictly separates SuperAdmin, Moderator, and Member permissions so administrative actions cannot be forged by untrusted socket clients.

### ERD Architect AI (AI Developer Tool)
*Instant Entity-Relationship Diagram generator parsing SQL, Prisma schemas, and natural language prompts.*
- **Stack**: Next.js 15, ReactFlow, Google Gemini 2.5, ELKjs, TypeScript, Tailwind CSS v4
- **Live Demo**: https://ai-er-diagram-generator.vercel.app
- **Source Code**: https://github.com/Sumit-y88/ai-er-diagram-generator

#### Architecture & Deep Dive
ERD Architect AI transforms raw SQL DDL, Prisma schema definitions, DBML, or plain English prompts into interactive, auto-layouted Entity-Relationship Diagrams in seconds. Built on Next.js and ReactFlow with Google Gemini 2.5 intelligence, it removes the friction of manual database diagramming.
> "Turn SQL, ORM schemas, and plain prompts into interactive, auto-layouted ER diagrams instantly."

**System Architecture:**
- **Client**: Next.js 15 App Router, ReactFlow (@xyflow/react), Tailwind CSS v4, Lucide React
- **REST API**: Next.js Route Handlers streaming Gemini 2.5 structural schema extraction
- **Realtime**: Client-side ELKjs layout calculation and interactive node position physics
- **Database**: Stateless diagram compiler with exportable JSON, SVG, and PNG serialization
- **Authentication**: Environment-isolated Gemini API key gateway with request sanitization

**Production Readiness & Engineering Hardening:**
- ELK (Eclipse Layout Kernel) auto-layout algorithms for clean Top-Down and Left-Right DAG structures
- Interactive relationship edge popups displaying foreign keys, cardinality, and constraints on hover
- High-resolution vector SVG and lossless raster PNG diagram export pipelines
- Multi-format schema input support (SQL CREATE TABLE, Prisma models, DBML, and conversational prompts)
- Responsive collapsible drawer layout optimized for both desktop workflows and mobile review

**Key Architectural Decisions:**
- **Automated Directed Graph Layout via ELKjs**: Eliminates overlapping edges and manual node positioning by running hierarchical layout algorithms directly in the browser.
- **Deterministic Schema Extraction with Gemini 2.5**: Prompts Gemini with strict JSON schema definitions to reliably map entities, attributes, primary keys, and relationships regardless of the input format.

### JobTrackr (Full-Stack MERN + AI Platform)
*AI-powered job application tracker with pipeline analytics, status stages, and cover letter generation.*
- **Stack**: React 18, Node.js, Express, MongoDB, TypeScript, Tailwind CSS, Google Gemini API, Framer Motion, Zustand
- **Live Demo**: https://ai-job-tracker-5zrn.vercel.app
- **Source Code**: https://github.com/Sumit-y88/Ai-job-tracker

#### Architecture & Deep Dive
JobTrackr is an AI-powered job application management platform designed to streamline the tech job hunt. Built with the MERN stack and TypeScript, it offers a Kanban-style pipeline (Saved → Applied → Interviewing → Offered → Rejected), interactive analytics charts, and automated cover letter generation via Gemini AI.
> "Streamline job hunt pipelines with automated AI resume matching and cover letter generation."

**System Architecture:**
- **Client**: React 18, Vite, TypeScript, Tailwind CSS, Framer Motion, Zustand, Recharts
- **REST API**: Node.js Express backend with RESTful resource routing and controller separation
- **Database**: MongoDB with Mongoose schemas for users, jobs, application stages, and cover letters
- **Authentication**: Dual auth: Email/Password (bcryptjs + JWT) and Google OAuth 2.0 via Passport.js

**Production Readiness & Engineering Hardening:**
- End-to-end type safety across models, API responses, and client forms via Zod
- Multi-stage pipeline tracking (Saved, Applied, Interviewing, Offered, Rejected) with persistent notes
- AI-driven cover letter generation tailored to candidate resume and job description
- Visual Recharts dashboard displaying application trends, status ratios, and weekly activities
- Security hardening with helmet, cors origin whitelisting, and express-rate-limit protection

**Key Architectural Decisions:**
- **Centralized Resume Hub Architecture**: Persists candidate resume profile once so all AI scoring and cover letter generation features execute instantly without re-uploading files.
- **Decoupled Client & Server Deployment**: Frontend deployed to Vercel with SPA rewrites; backend containerized and deployed to Render with MongoDB Atlas clustering.