Production hub with 35+ microservices powering the entire webtoon localization pipeline — from raw scan to published chapter.

• OCR, translation, segmentation, and ML vision pipelines
• Custom datasets with 30K+ annotated images
• Fine-tuned vision LLM for detection and formatting
• Active learning pipeline mining real production data

Community Discord bot serving thousands of users with character gacha, automated quizzes, virtual economy, and AI-powered features.

• Character gacha with League of Legends + Comikey series characters
• Automated quiz engine with self-healing after crashes
• Multi-tier caching for sub-second response times
• 100+ slash commands with role-based permissions

Manga/comics reading platform with payments, community features, admin panel, and chapter processing pipeline.

• Manga reader with Cloudflare Workers image proxy
• Stripe payments + WebSocket messaging
• Chapter processing with image optimization
• Discord OAuth + NextAuth v5

Multi-language OCR hub with web editors, automatic PSD generation, and 5+ AI providers for manga translation workflows.

• Custom OCR pipeline with dedicated JP/KR/ZH handlers
• Auto-generates layered Photoshop files from scans
• Batch processing pipeline for entire volumes
• Multi-AI routing: Claude, Gemini, OpenAI, Mistral

End-to-end pipeline for detecting and segmenting text in manga/manhwa/webtoons.

• 13-class bubble detection + 2-class segmentation
• 5000+ synthetic annotated images
• Active learning pipeline mining real production data
• Instance segmentation with polygon masks

AI-powered legal assistant for Brazilian Labor Law. Queries 80+ courts in real-time and generates submission-ready documents.

• Autonomous agent that researches case law via DataJud API
• Auto-generates formatted DOCX ready for judicial submission
• Real-time progress tracking from upload to final document
• Multi-model fallback: Claude → GPT-5 → Gemini

Automated chapter release monitoring with cross-platform publishing and AI-powered community Q&A.

• Parallel scraping with headless browser pool
• Auto-publish announcements to Twitter, Instagram, Bluesky
• AI-powered Q&A trained on series knowledge base
• Web dashboard for content management

Discord bot that generates AI videos from images using Wan2.2 with custom LoRA models.

• Single + dual-frame modes, 480p to 1080p
• Custom LoRA adapters for accelerated generation
• Automatic CDN upload for large outputs
• Concurrent job management with cost tracking

Converts Clip Studio Paint files to Photoshop preserving text as editable vector layers — a feature unavailable in native CSP export.

• Standalone tool with zero external dependencies
• Full PSD/PSB support including large format files
• Preserves filters, gradients, and layer properties
• MIT licensed, used by digital artists worldwide

Production-grade bidirectional bridge syncing messages, threads, emojis, and files between Slack and Discord in real-time.

• Full threading and emoji mapping across platforms
• File/attachment forwarding with format conversion
• Fault-tolerant with automatic reconnection
• Running in production for the entire team

Discord bot for managing modded Minecraft servers with real-time monitoring and remote control.

• Live log streaming with automatic IP masking
• TPS, RAM, and player monitoring with alerts
• Full RCON command execution from Discord
• Auto-detection of server restarts

CPU-optimized pipeline that detects text, removes it via inpainting, and translates — all in one shot.

• RT-DETR detection + LaMa inpainting via ONNX
• Intelligent strip slicing to avoid cutting bubbles
• Gemini-powered translation with text rendering
• No GPU required — runs on any machine

Automated penetration testing with 27 specialized sub-agents covering the full security assessment lifecycle.

• Multi-phase: Recon → Vuln Analysis → Exploit → Report
• Injection, XSS, Auth bypass, IDOR testing
• Git-based checkpointing and audit logging
• Claude Agent SDK + Playwright automation

Conversational AI assistant with Discord integration, LangChain orchestration, and local LLM support via Ollama.

• LangChain agent with tool-calling capabilities
• Local inference with Ollama for privacy
• Discord bot interface for team access
• Dockerized deployment with CI/CD

Multi-method watermark detection system combining template matching, feature extraction, and saliency analysis.

• SIFT/ORB feature matching for robust detection
• Saliency-based patch generation
• Support for 100+ user-specific watermarks
• Training pipeline with data augmentation

Desktop tool for extracting high-resolution preview images from Clip Studio Paint files with optional AI upscaling.

• GUI with recursive directory scanning
• Multiple extraction modes: preview, full-res, bitmap
• waifu2x neural network upscaling (2x/4x)
• Optimized for manga and anime artwork