02 // Curated Works

Technical Projects & Architecture Archive

Chronologically curated machine learning systems, computer vision streaming pipelines, and client-side operating tools. Every system features full visual pipeline nodes, tensor specs, and hardware constraints.

RHCSA Trained Red Hat Enterprise Linux (RHEL 9) Administration & Automation

Storage: LVM/VDO · Security: SELinux Enforcing · Units: systemd custom targets · firewalld

2024/ml_accident_detection.py ●

2024 · Computer Vision / Binary Classifier

ML Accident Detection System

Frame-level classification pipeline analyzing live traffic surveillance streams to distinguish collision events from normal vehicle motion, achieving approximately 82% validation accuracy on test streams.

[Arch: CNN Spatial Preprocessing / Frame Buffer] [Target: Low-Latency CCTV Video Feeds] [Validation: ~82% Test Stream Accuracy]

Python OpenCV TensorFlow Scikit-Learn NumPy

Source Code & Documentation ↗ v1.4.2 [STABLE]

2023/object_detection_stream.py ●

2023 · Deep Learning / Edge Stream

Real-Time Object Detection Pipeline

High-throughput multi-class bounding-box tracking pipeline targeting edge devices and webcam inputs. Employs non-maximum suppression (NMS) to eliminate redundant overlaps and maintain 30+ FPS throughput.

[Arch: Single-Shot MultiBox / NMS Filter] [Target: Edge Camera & Video Stream I/O] [Validation: 30+ FPS Real-Time Inference]

Python TensorFlow OpenCV Edge I/O

Source Code & Documentation ↗ v2.0.1 [STABLE]

2026/shrinkos_engine.js ●

2026 · Client-Side Tooling & Web OS

ShrinkOS In-Browser Analytics & Lab

Zero-dependency multi-page web operating system. Features client-side confusion matrix metric calculators with dynamic proportion heatmap rendering, high-DPI activation function gradient graphers, and in-browser CSV statistical analyzers.

[Arch: Vanilla HTML5 / CSS3 / ES6+ Canvas] [Target: Client-Side Edge Execution] [Validation: Zero Dependencies / 100% Offline]

HTML5 Canvas Vanilla CSS3 Modern ES6+ FileReader API

Launch ML Lab ↗ v2.4.0 [ACTIVE]