← Projects
Internal Tool · NLP · AI

Robotina

Meeting intelligence system that turns conversations into actionable data. Multi-device recording, automatic transcription, speaker identification by voice profile, and AI-generated summaries — all in a single internal tool.

The Problem

Meetings generate information that nobody captures

In any organization, the most important decisions are made in meetings. But most of those conversations end without a structured record: who said what, what was decided, what tasks came out of it. Information gets lost between incomplete notes and imperfect memories.

Robotina exists to solve that. It's not a bot that joins your Zoom — it's a system that records, transcribes, identifies who is speaking, and generates an intelligent summary with decisions and tasks. Fully automatic, fully private.

The Pipeline

From raw audio to meeting intelligence

01

Multi-Device Recording

Multiple devices can record the same meeting simultaneously. Audio is uploaded in chunks and merged in a unified pipeline. Works in any browser — no apps, no plugins.

02

Transcription with Diarization

Audio is sent to AssemblyAI for automatic transcription with diarization — separating who said what. Each segment includes precise word-level timestamps.

03

Speaker Identification

Each employee can register their voice profile. Pyannote compares the voices in the audio with registered profiles, assigning real names to each transcription segment instead of “Speaker 1, Speaker 2”.

04

Intelligent Summary

A language model generates a concise summary, extracts key decisions and lists action items with assigned owners. It responds in the same language as the conversation. All in structured JSON, ready for integration.

Organizational Context

Summaries that understand your organization

Robotina doesn't generate generic summaries. Each organization can configure its context: industry, internal glossary, departments, active projects and current priorities. The AI model uses this context to produce summaries that speak your team's language.

If your team says “the V2 of the pipeline”, Robotina knows exactly which project they're referring to. If they mention “Marcos' deadline”, the system associates that task with the right person.

The system understands your meetings because it understands your company.

Features

What comes out of every meeting

Full Transcription

Full meeting text with timestamps, speaker identification by name, and confidence levels per segment. Navigable and searchable.

Executive Summary

2–4 sentences that capture the essence of the meeting. Not a list of everything that was said — but what mattered.

Action Items & Decisions

Tasks automatically extracted with assigned owners when mentioned. Key decisions listed separately. Everything ready to move to your project management system.

Voice Profiles

Each team member registers their voice once. From then on, the system automatically recognizes who is speaking in each meeting, without the need for manual tagging.

Meeting Categories

Organize meetings by type: standup, planning, 1-on-1, client. Each category can have its own summary template and its own processing rules.

Organization Dashboard

Each organization has its private space with meeting history, employee directory, voice profiles and context configuration. Multi-tenant from day one.

Stack

How it's built

Frontend

Next.js + TypeScript. PWA with in-browser recording and offline support.

Backend

Supabase (PostgreSQL + Auth + Storage). API routes in Next.js with webhooks for asynchronous processing.

Transcription

AssemblyAI for speech-to-text with diarization. Pyannote for speaker identification by voice profile.

AI

OpenAI for summary generation with injected organizational context. Output in structured JSON.

Interested in Robotina?

Robotina is an internal tool we are actively developing. If your organization needs something similar, let's talk.

Contact Us