Luxury fashion and jewellery e-commerce platform, live in production at infiloft.com with real customer payments, not a portfolio demo.
- Architected and deployed the full stack on a self-managed Hostinger VPS: Django backend, two Next.js applications (customer storefront and admin panel), PostgreSQL, Redis, Nginx, Gunicorn, and PM2, with SSL and DNS fully configured end to end.
- Implemented a JazzCash/EasyPaisa manual-transfer payment flow with cash-on-delivery priority, tailored to real local buyer behavior, plus Google OAuth for customer authentication.
- Sole engineer: architecture, backend, both frontends, infrastructure, and ongoing operations.
DjangoNext.js × 2PostgreSQLRedisNginxGunicornPM2Google OAuth
02 / Agentic AI · Automation
AI Employee OS
GitHub ↗
A full agentic business-operations platform built in a 14-day sprint, designed to hand small businesses a digital employee for CRM, invoicing, and communication.
- Multi-persona AI employees that parse natural-language instructions via Groq Llama 3.3 70B.
- Generate PDF quotations and invoices with a transactional rollback pattern.
- Integrate Gmail, WhatsApp (Meta Cloud API), and Google Calendar, transcribe voice notes with AssemblyAI, and OCR documents with tesseract.js.
- Scheduled reminders run through n8n, all behind an authenticated Next.js dashboard with MFA.
Node.js / ExpressNext.jsPostgreSQLGroq Llama 3.3WhatsApp Cloud APIAssemblyAIn8nStripe
03 / AI Automation · Client Work
Clinic Lead Generation System
A self-hosted automation pipeline that fully automates discovery and qualification of 100+ clinic leads per week, without manual outreach.
- Self-hosted n8n on Oracle Cloud, orchestrating discovery through the Google Maps/Places API.
- Groq AI classifies and qualifies each lead automatically.
- WhatsApp Business Cloud API verifies phone numbers before a lead reaches staff, with results logged to Google Sheets.
- Two-person build: I owned the AI classification agent and the WhatsApp verification layer.
n8nGroqGoogle Maps/Places APIWhatsApp Business APIOracle Cloud
04 / Full-Stack · Client Work
Gym Management System
A full gym management application built for Body Reshaping Studio, a real client, running live on-site on the gym's Windows PC.
- Integrated a ZKTeco K50 biometric fingerprint device via a custom Python (pyzk) agent running as an NSSM Windows service, streaming attendance events in real time to the web app over Socket.IO.
- Solved a persistent device-connectivity and COM SDK debugging challenge to get reliable fingerprint enrollment working end to end in production.
- Sole engineer: hardware integration, backend, and frontend.
ReactNode.js / ExpressPostgreSQLPrismaSocket.IOZKTeco pyzk
05 / Computer Vision · Final Year Project
AI-Powered Video Surveillance & Analytics System
A real-time multi-camera surveillance pipeline for SZABIST Islamabad's final year project, built solo.
- YOLOv8 for detection and DeepSORT for multi-object tracking, streamed over WebRTC and served through a Flask backend, running at 15 to 25 FPS on CPU.
- Event-driven alerting layer that flags anomalous activity automatically, reducing dependence on continuous manual monitoring.
- Revised the full technical report systematically to close gaps identified in supervisor feedback, tightening methodology and evaluation sections.
YOLOv8DeepSORTOpenCVFlaskWebRTC
Abdul Samad