NeuroSort AI
A privacy-first desktop application that uses a local Phi-3 model to intelligently organize your chaotic Downloads folder into a structured Twin-Directory system.
→ index.ts [Regex Match]
✓ Moved to ./00_Source
→ img_921.raw [AI Reason]
✓ Moved to ./05_Assets
Reason: \"File contains heavy metadata associated with professional photography workflows...\"
The Challenge
The typical user's Downloads folder is a digital wasteland where productivity goes to die. Organizing it manually is a soul-crushing chore, yet handing this data to cloud providers for classification means sacrificing absolute privacy for a cleaner hard drive.
Market Insight
With the rise of local LLMs, the need for "Privacy-First" productivity tools has skyrocketed. NeuroSort leverages this paradigm shift, proving that state-of-the-art AI doesn't need a server connection to create immense personal value.
Key Outcomes
The 3-Stage Funnel
Stage 1: Regex Guard
An ultra-fast pattern matching system that handles ~70% of files instantly based on extensions and common naming conventions.
Stage 2: Metadata Analyst
Inspects file headers and EXIF data to classify media and documents, handling an additional ~20% of files without AI overhead.
Stage 3: AI Arbiter
The final fallback using Phi-3 (3.8B LLM) via Ollama to reason through ambiguous files, ensuring 100% classification accuracy.