ArchiveofWork
A comprehensive collection of intelligent systems, automation APIs, and generative models I've engineered.
SelectedWork
TriState Ride - Full-Stack Luxury Chauffeur Platform (Passenger, Driver & Admin Ecosystem)
Passenger App: Expo (React Native) | Driver App: Expo (React Native) | Super Admin Panel: Next.js + Tailwind CSS | Backend/API: Node.js + Express.js | Database: PostgreSQL / MongoDB | Real-time: Socket.io | Maps: Google Maps API | Auth: JWT + OAuth | Cloud: AWS / GCP | Payments: Stripe
+6Built and deployed TriState Ride as a complete ecosystem of three connected applications: Passenger App (Expo), Driver/Chauffeur App (Expo), and Super Admin Panel (Next.js). The platform supports on-demand and scheduled bookings, vehicle-class selection, real-time chauffeur tracking, automated assignment, airport transfers, hourly bookings, and cross-state operations across New York, New Jersey, and Connecticut.
Click to view details →
Rolac World Cup Photobooth — AI Football Jersey Generator
Engineered a complete end-to-end system, not just a model demo. A React 18 + Vite SPA handles camera capture, team selection (10 countries) and result download; a Flask API served by waitress (64 threads) behind nginx generates the composite. Google Gemini fuses the selfie and a jersey reference in one pass, backed by a self-hosted CV pipeline (InsightFace buffalo_l + inswapper, ONNX Runtime, MediaPipe segmentation, GFPGAN/OpenCV) with affine alignment, LAB colour matching and seamless blending for direct face/head-swap modes. An admin panel reviews every generated record.
+8Click to view details →
Face Swiper — Production-Grade AI Face & Head Swap
Built a two-mode pipeline. FACE mode swaps the face while preserving the subject's hairline and surrounding context; HEAD mode performs a full-head transfer. The system uses InsightFace (buffalo_l detector + inswapper) on ONNX Runtime for the core swap, MediaPipe segmentation to isolate regions, affine alignment for geometry, LAB colour matching for tone, and seamless blending to remove seams. GFPGAN restores and upscales facial detail in the final output. Supports group photos and a UI face picker.
Click to view details →
V-TRYON AI — Virtual Clothing Try-On Studio
Built a two-stage AI pipeline. Stage 1 uses SegFormer B2 Clothes to segment upper-body clothing regions (shirt, top, jacket, dress) from the person photo and produce a dilated binary mask. Stage 2 feeds that mask to Stable Diffusion Inpainting with a carefully crafted fashion-photography prompt to generate the new garment realistically onto the body. The person image is resized to 512×768 (Stable Diffusion's sweet spot) and the result is returned as a base64 PNG. A modern dark glassmorphism UI offers drag-and-drop uploads, real-time API health status, progress indicators and one-click downloads.
Click to view details →
PinVault — Secure Internal Vendor Map Platform
Built a full-stack platform with a Node.js + Express + MySQL backend and a fast single-file HTML/Leaflet frontend. Role-based access (super admin vs employee) is enforced with JWT + bcrypt. Vendors can be created from a location search or a direct map click, with fields for description, notes, phone, website, category and coordinates. Super admins manage staff, edit/remove vendors, and import/export data in CSV and JSON, plus export the map as an image. Browser-side privacy deterrents include username/timestamp watermarking, right-click and copy blocking, dev-tools deterrence, and Print-Screen detection with a capture-shield overlay.
+2Click to view details →
Lifebuoy FANFEST — Player Pack AI Photobooth
Built a Flask service with an async pipeline: upload + jersey index → Gemini dresses the person → rembg cutout composited on a themed background → fit into the transparent window of the matching Player Pack frame → return the finished image. Generation is asynchronous — /generate returns a job_id instantly and booths poll /status/<id> (queued → generating → removing_background → framing → done), with queue_ahead so a fan sees "you're #3 in line". The background-removal model loads once at startup, frames are cached/downscaled per size, and uploads are capped to 1280px. The prompt keeps a hijab/headscarf and modest outfit when the person wears one. Includes an admin dashboard to review and delete generated packs.
Click to view details →
Ruchi Photo Booth — Stand Beside Your Star Player (AI)
Built a FastAPI backend where the /photobooth route posts a country + base64 image and gets back a generated composite plus the player's name. Gemini generates the whole scene — the fan next to the player in that stadium, both wearing the jersey — while keeping the fan's real face and identity (no face editing on our side). Country/player assets and jerseys are matched by country prefix with aliases; a branding overlay is composited on top. Model order favours gemini-3-pro-image-preview for the best jersey dressing and falls back to gemini-2.5-flash-image on overload.
Click to view details →
Wafid Auto-Fill — Passport OCR Chrome Extension
Built a Chrome extension + FastAPI OCR backend. The user uploads a passport (image or PDF); the backend runs pytesseract OCR with a dedicated MRZ pass that acts as the reliable source of truth for name, DOB, passport number and expiry, plus opportunistic page extraction for email/phone/national ID. Manual fields (country, city, visa type, etc.) are saved in the popup and persisted across sessions. Anything OCR misses is flagged with a ⚠️ badge for one-time inline editing, then bundled into the auto-fill payload — no re-OCR needed. The content script fills the live wafid form using configurable selectors.
Click to view details →
Page 1 of 4 · 52 projects
Let'sBuildSomethingIntelligentTogether
Available for full-time roles, freelance projects, and technical consulting engagements.
- mdzubayerhossainpatowari@gmail.com
- +880 1841 606311
- Dhaka, Bangladesh