Software Engineer & Maker · Building pragmatic, high-performance, local-first software across Applied AI & MCP, Native macOS (Swift), Audio DSP (C++/JUCE), and Self-Hosted Systems.

English · 🇮🇹 Italiano

• EngramVault — Self-hosted, local-first AI Memory Vault, Epistemic Knowledge Graph (FACT / CLAIM / DECISION), Hybrid Search (FTS5 BM25 + 384-dim semantic vectors + RRF), and native Model Context Protocol (MCP 2.0) server for AI agents (Antigravity, Claude Code, Cursor).
  Stack: Python, SQLite (WAL & FTS5), FastAPI, MCP 2.0, PWA

• CozyTabs — Personal Telegram bot using multimodal AI (Gemini audio + video reasoning) to analyze guitar video lessons (Reels, TikTok, Shorts) and generate structured study cards with ASCII tablatures and chords on Notion.
  Stack: Python, Google Gemini Multimodal, Notion API, Telegram Bot API

• AiSentry — Native macOS menu bar utility for monitoring real-time consumption, quotas, and session limits across multiple AI providers (Claude, OpenAI, Google Gemini, Grok) with secure Apple Keychain storage.
  Stack: Swift, SwiftUI, macOS MenuBarExtra, Keychain Services

• PingSentry — Lightweight native macOS menu bar network monitor displaying real-time latency and packet loss using Wi-Fi-style indicator bars.
  Stack: Swift, SwiftUI, Network.framework

• Livellatore — Dynamic voice auto-leveling VST3/AU audio plugin inspired by Sonic Anomaly TriLeveler, featuring target-level leveling, breath gate, and transparent brickwall limiting for podcasting and voice-overs.
  Stack: C++, JUCE, Audio DSP

• Cozy-Bookmarks — Minimalist self-hosted bookmark manager featuring clean Material 3 design, zero-build-step frontend, automated multi-provider AI tagging, strict anti-SSRF validation, and Android PWA support.
  Stack: FastAPI, HTMX, SQLite, Material 3, Docker

Languages      │ Swift · C++ · Python · TypeScript · SQL · Bash
Audio & DSP    │ JUCE Framework · VST3 / AU · Audio Signal Processing
AI & Agents    │ Model Context Protocol (MCP) · Multimodal LLMs · RAG / Vector Search · FTS5
Backend & Web  │ FastAPI · HTMX · SQLite (WAL/ACID) · REST APIs · Server-Sent Events
Platforms      │ macOS Native · Linux / Homelab · Docker · Embedded / IoT

Sviluppatore software e maker appassionato di ingegneria dei sistemi, elaborazione del segnale e intelligenza artificiale applicata.

Mi occupo di progettare software concreto, veloce e orientato al controllo dei dati:

• AI Applicata & Protocolli Agentici: implementazione di server MCP nativi, grafi di conoscenza epistemici e pipeline multimodali (audio/video).
• Utility macOS Native: strumenti puliti in Swift/SwiftUI che risiedono discretamente nella menu bar con impatto minimo sulla memoria.
• DSP Audio: algoritmi ed elaborazione numerica dei segnali in C++ per l'elaborazione dinamica della voce e la modellazione audio analogica.
• Self-Hosted & Homelab: architetture containerizzate robuste, incentrate su SQLite, zero build-step e sicurezza rigorosa (anti-SSRF, session hashing).

• GitHub: @gaskb
• Email: matteo.gasparri@gmail.com
• Support: Se trovi utili i miei progetti open-source, puoi offrirmi un caffè su Ko-fi ☕.