Computer Engineering (Hons) graduate from UTeM, available immediately. Building end-to-end AI systems — from YOLOv8 real-time vision to LLM multi-model agent orchestration and data pipelines. Cisco CCNA certified, Festo Industrial AI specialist. Currently completing IBM Data Engineering on Coursera. Open to relocation.
I build systems that think. My philosophy: Build, Test, Break, Understand, Improve. From YOLOv8 anti-fraud vision at checkout counters to a 13-model consensus system on a phone running Termux. Available immediately for full-time roles in AI engineering, networking, or industrial automation.
Designed to prevent item evasion and fraud in automated self-checkout systems by cross-referencing visual YOLOv8 object detection against physical barcode scans in real-time.
An AI agent running on a Samsung Galaxy A54 (Termux/Linux, aarch64, 7.3 GB RAM). Two processes: hermes serve (~170 MB RSS) and hermes gateway (~280 MB RSS), restarted every 2 minutes by a cron watchdog. Queries 13 LLM models (DeepSeek, GLM, Kimi, Minimax, Hunyuan) through a Parliament of Minds consensus engine for XAUUSD trade signals. Includes a causal post-mortem system (6 analyses to date) and Google Workspace integration (8 OAuth scopes). 103 closed signals from Aug 24 to Sep 15 2026; system paused since Sep 22 strategy refit.
An interactive web mini arcade featuring intelligent AI Tic Tac Toe with minimax decision tree logic, soccer penalty shootout mechanics, and sports card memory match. Deployed on Cloudflare edge under custom subdomain with ultra-responsive gameplay.
Zero-touch reading assistant for PC and mobile webtoons to read while eating without touching greasy screens. Powered by proprietary 71D hand geometry feature engineering and a 100-tree Random Forest classifier trained directly on real video demonstrations, achieving 98.6% cross-validation accuracy, zero-recoil geometric classification (eliminating return-stroke false triggers), 0.30s debounced scrolling, and virtual air buttons.