Hi, I'm Abdullah Altaf
Software Engineer|React Native & Full-Stack Specialist|AI & Computer Vision
I bridge production-grade AI and computer vision with seamless mobile and web experiences — from a tap on a phone screen down to a quantized model running on edge hardware.
Academic foundation
The formal grounding behind the applied AI and engineering work.
BS Computer Science — AI Specialization
3.82 CGPAPMAS-Arid Agriculture University, Rawalpindi
2022 — 2026
Where I've put this to work
Hands-on roles alongside the engineering work — including teaching the fundamentals I apply daily.
React Native Full-Stack App Developer
Full-Time · On-siteSoft Wise Solutions, Rawalpindi
Jan 2025 — Present
Computer Science & AI Instructor
Part-Time · On-siteUnique Future Academy, Rawalpindi
Oct 2025 — Present
An ecosystem built for mobile, backend, and edge intelligence
Each category represents a layer of the stack I own end-to-end — from a tap on a phone screen down to a quantized model running on a Raspberry Pi.
Mobile & Frontend Engineering
- React Native
- React.js (MERN)
- Redux Toolkit
- Zustand
- TanStack Query
- Axios
- React Navigation
- MVC Architecture
- Clean Architecture
Backend & Database Infrastructure
- Node.js
- Express.js
- Prisma ORM
- MongoDB
- RESTful APIs
- Secure Middleware
- Rate Limiting
- RBAC
AI, Computer Vision & Edge Intelligence
- YOLO
- PaddleOCR (Mobile & Server)
- Custom Dataset Collection
- Web Scraping (Python)
- Beautiful Soup / Scrapy
- CVAT
- LabelOCR
Performance & Optimization Core
Edge & OCR Optimization
Server-grade accuracy on-deviceCross-compiling and fine-tuning lightweight Mobile PaddleOCR models, optimizing quantization and inference pipelines to match large server-model accuracy constraints in high-throughput environments.
Mobile App Performance
Zero JS-thread blockingReducing React Native JS-thread blockages using MMKV for lightning-fast synchronous storage reads/writes, implementing Zustand for lean global state management, and asset caching.
Network & State Optimization
Resilient request lifecycleLeveraging TanStack Query for robust server-state caching, automatic background revalidation, stale-data management, and custom Axios interceptors for resilient request lifecycle management.
Database Query Optimization
Minimized round-tripsArchitectural optimization via Prisma ORM for MongoDB, structuring custom aggregations, and query indexing to minimize database roundtrips.
SmartPark
An AI-IoT driven edge parking ecosystem with automated annotation and multi-role enterprise management — spanning a Raspberry Pi vision pipeline to a full RBAC web platform.
Event-Driven Resource Optimization
The edge device (Raspberry Pi) stays idle until physical sensors detect a vehicle, triggering an interrupt loop that captures exactly 3 frames — avoiding continuous compute burst overhead.
Dual-Stage Vision Pipeline
Frames pass through a custom-trained YOLO model to isolate license plate coordinates. The pipeline dynamically crops the bounding box to remove background noise before passing the refined segment to an optimized Mobile PaddleOCR variant.
Hardware Actuation & Synchronization
Upon server verification, the Raspberry Pi fires GPIO signals to a relay module, engaging a heavy-duty modified wiper motor mechanism to actuate a physical barrier with a 10-second cool-down auto-close cycle.
MLOps Pipeline Highlight
Custom YOLO checkpoints integrated directly inside CVAT to automate raw image pre-labeling, with LabelOCR used for custom training data generation to fine-tune PaddleOCR for non-standard local license plate formats.
Global School System
A cross-platform school management app built with React Native — giving students and faculty a single portal for academics, attendance, fees, and scheduling.
Multi-Role Portals
Dedicated dashboards for students and faculty — student ID cards, teacher class rosters, and role-aware navigation from a single codebase.
Academic Records Engine
Term-wise results with subject-level breakdowns, auto-computed percentages and grades, and a running datesheet with live status (completed / ongoing / scheduled).
Attendance & Fee Operations
Calendar-based attendance tracking with present/absent/late/holiday tallies, plus itemized fee ledgers with receipt numbers and payment history.
Qureshi Electronics — Inventory & Stock Tracker
A role-scoped inventory and point-of-sale platform for a multi-branch electronics retailer — one codebase serving four permission tiers, from full cross-branch oversight down to a salesman's read-only product view.
Branch-Scoped RBAC
A single codebase serving four distinct roles, with row-level data scoping so a Branch Admin only ever queries their own shop's records.
Consolidated Financials
Real-time gross profit, net profit, tax paid, and commission roll-ups — aggregated across shops for Super Admin, or filtered per-branch.
Inventory & Purchase Tracking
Warehouse-aware product availability, consolidated purchase invoices, and searchable stock ledgers by product code, type, or company.
Role-Based Access Control
Super Admin
Consolidated visibility across every shop branch — cross-branch profit, purchases, expenses, and commission reporting from a single dashboard.
Branch Admin
Scoped strictly to one branch — manages that shop's inventory, invoices, deposits, and expenses without visibility into other branches' data.
Stock Manager
Owns all stock-related tasks — purchase invoices, consolidated stock counts, and warehouse-level product availability across the branch.
Salesman
Restricted to a read-only view of available products for sale, with no access to purchase, financial, or cross-branch data.
Start a conversation about your next build
Whether it's a React Native app, a full-stack platform, or a computer vision pipeline that needs to ship on real hardware, I'm glad to discuss scope, architecture, and timelines.
Abdullah Altaf
React Native · Full-Stack · Edge AI & Computer Vision