---

About Me

const shahnawaz = {
  name:     "Shahnawaz Shaikh",
  location: "Mumbai, India",
  role:     "Full Stack Engineer & AI Builder",

  currentlyBuilding: [
    "Portfolio Forge  — AI resume to portfolio generator",
    "Real-time collab — WebSockets + CRDT state sync",
    "AI Code Reviewer — LLM-powered PR review bot",
    "Headless UI Lib  — Tailwind + TypeScript, zero JS",
  ],

  stack: {
    ai:       ["Groq", "Gemini", "OpenAI", "LangChain"],
    backend:  ["Node.js", "FastAPI", "GraphQL", "Redis"],
    frontend: ["React", "Next.js", "Tailwind", "Vite"],
    devops:   ["Docker", "GitHub Actions", "CI/CD"],
  },

  funFact: "I turn resumes into portfolios with AI",
  openTo:  "Collabs, Freelance, Full-time roles",
};

---

Tech Stack

Languages

         

Frontend

         

Backend

         

Databases & Cache

         

AI / ML

         

DevOps & Cloud

         

Auth · Testing · Tools

                   

---

Featured Project — Portfolio Forge

Upload a resume. Get a production-ready portfolio website in seconds. 100% browser-based. Zero backend.

Feature	Details
Multi-Provider AI	Groq (Llama 3.3 70B) · Gemini 2.0 Flash · OpenAI GPT-4o Mini · OpenRouter (Mistral 7B)
Key Pool Rotation	Add multiple API keys — auto-rotates on rate-limit, never blocks you
In-Browser Parsing	PDF.js + Mammoth.js — extracts text 100% in the browser, zero server uploads
6 Design Presets	Dark Minimal · Editorial · Terminal · Glassmorphism · Bold · Clean
One-Click Export	Downloads a clean ZIP with separate index.html, style.css, script.js
36 Unit Tests	Vitest suite — parsers, validators, file handling, typed error classes
Auto-Deploy	GitHub Actions CI/CD → GitHub Pages on every main push
Privacy-First	Keys in JS memory only — never localStorage, never a server

Architecture Deep-Dive (click to expand)

portfolioforge/
│
├── index.html                # App shell: semantic HTML + ARIA roles, zero inline styles
├── src/
│   ├── style.css             # Design system via CSS custom properties (20+ variables)
│   └── app.js                # All logic — vanilla ES2022, zero runtime dependencies
│       ├── STATE {}           # Single source of truth: provider, keyPool[], html/css/js
│       ├── PROVIDERS {}       # Per-provider config: endpoint, model, validate()
│       ├── Key Pool Engine    # addKeyToPool() → markKeyRateLimited() → getNextKey()
│       ├── callWithKeyRotation()  # Round-robin over keys → multi-round backoff
│       ├── extractPDF()       # PDF.js page-by-page Y-coordinate text reconstruction
│       ├── extractDOCX()      # Mammoth.js raw text extraction
│       ├── cleanResume()      # Normalise whitespace, hard-cap at 1,800 chars
│       ├── callGemini()       # Gemini-specific request format + error mapping
│       ├── parseAI()          # 4-strategy: delimiters → fences → normalised → raw HTML
│       └── renderPreview()    # Sandboxed srcdoc iframe with injected CSS + JS
│
├── tests/app.test.js         # 36 tests across 6 describe blocks
└── .github/workflows/ci.yml  # test → validate → deploy pipeline

---

GitHub Stats

   

---

What This Repo Actually Demonstrates

Every claim below is directly evidenced by code in this repository.

Signal	Evidence in Code	Depth
AI / Multi-Provider	4 LLM providers, unified abstraction, prompt engineering, token budgeting	█████████░
API Integration	REST calls to 4 different provider APIs, correct headers, per-provider error mapping	████████░░
Key Pool & Rate Limiting	Round-robin rotation, backoff, markKeyRateLimited(), multi-round retry	█████████░
In-Browser File Parsing	PDF.js Y-coordinate reconstruction, Mammoth.js DOCX extraction	████████░░
State Management	Vanilla STATE {} pattern — single source of truth, no framework needed	████████░░
Testing	36-test Vitest suite: validators, parsers, file handling, typed error classes	████████░░
CI/CD	GitHub Actions: test → HTML validate → deploy to Pages on every push	█████████░
Privacy / Security	Keys in JS memory only — never localStorage, never sent to a server	█████████░
CSS Architecture	Design system via 20+ CSS custom properties, 6 distinct design presets	████████░░
Error Handling	Custom RateLimitError + ParseError classes, 4-strategy HTML parser	████████░░

Note: Auth, databases, Docker, TypeScript, WebSockets, and other skills listed in the Tech Stack badge section reflect my broader experience and other projects — not this specific repo, which is intentionally zero-dependency and serverless by design.

---

Currently Building

Portfolio Forge v4     ──  Claude / Anthropic streaming + live generation view
Real-time collab tool  ──  WebSockets + CRDT-based conflict-free state sync
AI code reviewer       ──  GitHub Action that reviews PRs via LLM feedback
Open-source UI lib     ──  Headless Tailwind + TypeScript components, zero JS

---

Engineering Philosophy

"Ship it, then improve it.
 Tests first on anything that has state.
 If it needs a server, question why.
 If it needs a framework, question harder."

Principle	Practice
Simplicity over cleverness	Readable 10-line function beats a smart 3-liner
Ship early, iterate fast	Working v1 beats perfect v0 every time
Test behaviour, not implementation	Tests describe what, not how
Security by default	Keys in memory only · HTTPS always · no implicit trust
Zero-dependency where possible	Vanilla JS over a framework for focused tools
Docs are code	README, comments, and types are first-class citizens

---

Open to Opportunities

---

If any of my work helped you — a star means the world to me.