# Nicolò Tognoni

> Forward Deployed Engineer at Wonderful.ai. Based in Milan, Italy. Official portfolio: https://nicolotognoni.com/

This overview describes the public portfolio. For complete text, see /llms-full.txt. It is a supplemental resource; the HTML portfolio is the primary source.

## Portfolio sections

- [About](https://nicolotognoni.com/#about)
- [Experience](https://nicolotognoni.com/#career)
- [Projects](https://nicolotognoni.com/#work)
- [Education and publications](https://nicolotognoni.com/#writing)
- [Contact](https://nicolotognoni.com/#contact)

## Case studies

- [Patter: an open-source voice AI SDK](https://nicolotognoni.com/projects/patter): From 200+ calls with customers, developers and startups to an open-source voice AI SDK with 1,000+ GitHub stars, 126 forks and 9 contributors.
- [Prisma: an AI workspace across desktop, web and MCP](https://nicolotognoni.com/projects/prisma): A unified AI workspace that brings conversations, agent tools, email and calendar together through a shared backend and one interface for desktop and web.
- [Digital Twin: shared AI memory and collaboration through MCP](https://nicolotognoni.com/projects/digital-twin): A personal AI agent whose memories and plans live outside a single chat, with tools for collaborating with the Digital Twins of connected friends.
- [AI Council: multi-agent debate with a live ChatGPT interface](https://nicolotognoni.com/projects/ai-council): Five agents with distinct roles work through a question in structured rounds, while a live interface exposes their proposals, critiques and votes.

## Projects

- [Patter](https://github.com/PatterAI/Patter): An open-source SDK that gives AI agents a phone number — connecting any agent to real phone calls in 4 lines of code. Python and TypeScript at parity, MIT licensed, sub-500ms time-to-first-audio. Pluggable across 27+ providers including OpenAI Realtime, Anthropic, Deepgram, ElevenLabs, Twilio and Telnyx.
- [Prisma](https://nicolotognoni.com/#work): An Electron desktop app (TypeScript, React) providing a unified AI-driven workspace. The Node.js / TypeScript backend handles reasoning, state and orchestration; external services are wired through MCP servers (OAuth2, gRPC/HTTP, JSON-RPC) to expose email, calendar and other APIs in a secure, standardized way.
- [Digital Twin](https://github.com/nicolotognoni/DigitalTwin): First-prize winner at the MCP Apps hackathon in Turin, sponsored by Manufact. An AI-powered personal agent with a centralized memory layer. Create your Digital Twin, collaborate with friends' Twins via MCP, and access your plans and memories from any AI client — ChatGPT, Claude Code, or your own apps. Built with TypeScript and Supabase.
- [Wayfarer](https://github.com/nicolotognoni/Wayfarer): AI travel companion that talks, searches and calls restaurants for you. Voice chat to discover places, view interactive cards with photos and ratings, and let the AI make phone calls in 15 languages. Built with SwiftUI, ElevenLabs Conversational AI, Firecrawl, Google Places API, and Twilio.
- [AI Council](https://github.com/nicolotognoni/ai-council-mcp): An MCP server that orchestrates debates among five AI agents with distinct personalities — Strategist, Pragmatist, Critic, Innovator, Philosopher. Users pose complex questions, and the agents deliberate across structured rounds (Proposal, Critique, Refinement, Vote) before reaching consensus. Built with TypeScript, the mcp-use framework, and the OpenAI Apps SDK, with a live-streaming React widget for real-time visualization.
- [Flow](https://nicolotognoni.com/#work): An AI keyboard for iOS that lets users leverage AI directly from their iPhone keyboard. Create personalized messages, rewrite content, and auto-respond across any application.
- [Recapy](https://nicolotognoni.com/#work): An AI-powered app that turns hours of YouTube videos into minutes of actionable insights. Features comprehensive summaries, full transcripts, and key-quote extraction.
- [GAN for CNN Accuracy](https://github.com/nicolotognoni/GAN_for_increasing_CNN_accuracy): Research project exploring how GAN-generated synthetic data can improve CNN classification accuracy in low-data scenarios. Part of the International Master thesis.
- [Satellite Image Classification](https://github.com/nicolotognoni/Satellite_Image_Classification_CNN): A deep-learning project implementing Convolutional Neural Networks for satellite image classification, enabling automated land-use analysis.
- [Robotic Arm](https://nicolotognoni.com/#work): Currently suspended. The final vision: mount a camera on the arm and use AI image-recognition with motors driven via Arduino, so the AI controls the robot to play backgammon against a real person.

## Profiles

- https://www.linkedin.com/in/nicolotognoni/
- https://github.com/nicolotognoni
- https://x.com/Nicolo_Tognoni
- https://medium.com/@nicolotognoni