Benjamin De Kraker | Full Stack Developer

A collection of various projects and ideas. A work in progress.

OpenClaudeCode

OpenClaudeCode is a simple yet powerful Python tool that brings Anthropic's Claude AI to your terminal. Think of it as your personal coding assistant! Ask questions, run commands, manage files, or even handle Git tasks—all from a command-line interface (CLI) or an interactive REPL.

Key Features: Easy installation with git clone and pip install, configuration via a config.json file, and versatile usage through both quick CLI commands and an interactive REPL.

View on GitHub

Audio Analysis and Normalization Software

This Python application is a desktop tool for visualizing, analyzing, and processing WAV audio files. The application assisted my team on the training pipeline for Grok Voice.

It leverages several technologies and libraries:


PyQt5: Provides the graphical user interface (GUI), including menus, custom widgets, and media playback controls.
PyQtGraph: Used for plotting and dynamically updating waveform visualizations.
Pydub: Handles audio file manipulation such as normalization, gain adjustments, and conversion.
NumPy & wave: Perform low-level audio processing tasks like computing RMS (Root Mean Square) and peak levels.
FFmpeg (via subprocess): Ensures precise audio conversions (e.g., converting files to mono, 24-bit, 48 kHz).

Manual Volume Normalization: Implements a multi-pass, chunk-based normalization process with crossfading to smooth out transitions.
Audio Conversion: Provides a tool to convert audio files to the required specifications using both pydub and FFmpeg.

Read More

AI Vision with Object Recognition

I built a clone of a Google DeepMind demo -- it took under an hour.
Object recognition and description from video frames via GPT-4o Vision API.

Read More

AI Video Creation and Editing Platform

I was the founder and sole developer of FinalFrame.ai, one of the first consumer AI video platforms.

Tech Stack: Javascript, PHP, Python, MySQL, Linux, Redis
Backend: AI libraries on distributed NVIDIA A100 GPU machines, CUDA, Docker, Conda, custom Python

Read More

Lat / Lon GPS Coordinates for ChatGPT

Locate Me, GPT: Ping your location (anonymously) for Maps and Directions in GPT.

MVP / Demo of a secure method of doing tours, navigation, and location-based features for ChatGPT users.

Read More

Eye Tracking: In-browser Gaze Based Menu Navigation

Web application made with JavaScript. Handsfree face and eye tracking interface, such as for disabled users.

Read More

Terminal AI: Talk to your Terminal with AI + Speech Recognition

Windows/Mac application made with Electron + Node.js.

This is a fully functional custom terminal (CLI) for Windows, integrated with AI functions and advanced voice recognition.

Read More

FSD Stop Detection Dashcam Video Editor

Automatically detects stop/start events in dashcam videos by analyzing a visible speedometer. Using OCR and frame processing, it logs when the vehicle stops or moves, overlays speed and status on video, and exports clips of moving segments.

Web Version Here

Gesture Detection and Emoji Animation

A real-time system that uses MediaPipe for hand and face tracking to detect gestures (thumbs up, victory, facepalm, heart, OK hand) and displays floating emojis using OpenCV and Pillow.

Read More

Magnifier Lens Effect

A JavaScript-based magnifying lens effect for images, featuring smooth magnification that follows the mouse cursor and adjustable lens size via the scroll wheel.

View on GitHub

LLM Chain Builder with Nodes-based Logic

A Python application built with PySide6 that provides a graphical, node-based interface for designing and executing language model chains. The project utilizes Qt’s QGraphicsScene and QThread to allow users to add various node types—such as user input, code execution, and image nodes—and connect them to form complex workflows.

Tech Stack: Python, PySide6 (Qt for Python), QGraphicsScene, QThread, subprocess, and integration with APIs (e.g., Anthropics API) for LLM processing.

The code demonstrates how nodes are created, connected, and executed in sequence. It also includes features for saving and loading node configurations using JSON.

Read More