MeetGeek

Captures video and audio streams from calendar-integrated meetings across platforms (Zoom, Google Meet, Microsoft Teams, etc.) by hooking into the meeting application's media pipeline or using browser-based WebRTC interception. The system maintains persistent connection to the meeting session and buffers raw media streams locally or to cloud storage with automatic format conversion to standard codecs for downstream processing.

Converts recorded audio streams into timestamped text transcripts using automatic speech recognition (ASR) models, with speaker diarization to attribute each spoken segment to the correct participant. The system likely uses a multi-stage pipeline: audio preprocessing (noise reduction, normalization), ASR inference (possibly using Whisper, Google Speech-to-Text, or proprietary models), and speaker identification via voice embeddings or meeting metadata (participant list matching).

Integrates MeetGeek with multiple meeting platforms (Zoom, Google Meet, Microsoft Teams, Webex) using platform-specific APIs and bot frameworks. The system handles OAuth authentication, bot lifecycle management (joining/leaving meetings), and platform-specific features (Zoom recording API, Teams side panel integration, Google Meet activity tracking).

Analyzes full meeting transcripts to identify and extract the most important segments, decisions, and action items using a combination of extractive summarization (selecting important sentences from the original transcript) and abstractive techniques (generating concise summaries). The system likely uses NLP models to score sentences by relevance, detect decision-making language patterns, and identify action items via dependency parsing or sequence labeling, then ranks and presents results in a structured format.

Automatically extracts and structures meeting metadata including participants, duration, topics discussed, decisions made, and action items into a queryable database. The system parses calendar event data, transcript content, and summary outputs to populate a structured schema, then indexes this data for full-text search and faceted filtering. This enables downstream search and retrieval capabilities.

Monitors calendar systems (Google Calendar, Outlook, etc.) for scheduled meetings and automatically enrolls the MeetGeek agent in those meetings to begin recording and processing. The system uses calendar API webhooks or polling to detect new events, validates meeting type (excludes personal/blocked time), and injects the agent into the meeting session using platform-specific APIs (Zoom bot API, Teams bot framework, Google Meet API).

Provides live, streaming transcription and real-time insights during active meetings by processing audio in near-real-time (10-30 second latency) and displaying transcripts and key points to participants. The system uses streaming ASR APIs, incremental summarization, and live speaker diarization to update the transcript and insights as the meeting progresses, typically displayed via a web interface or meeting platform integration (Teams/Zoom side panel).

Enables full-text and semantic search across the entire meeting archive by indexing transcripts, summaries, and metadata, and using vector embeddings to find semantically similar meetings or segments. The system likely uses a combination of traditional full-text search (Elasticsearch or similar) for keyword matching and vector search (embeddings-based retrieval) for semantic queries, allowing users to find meetings by topic, decision, or action item rather than just keyword matching.

Automatically extracts action items from meeting transcripts and summaries, assigns them to participants based on explicit or inferred responsibility, and tracks completion status. The system uses NLP to detect action item language patterns ('I will', 'we need to', 'assigned to'), links action items to specific speakers, and integrates with task management systems (Jira, Asana, Microsoft To Do) to create and track tasks.

Manages storage, retention, and lifecycle of meeting recordings and associated data (transcripts, summaries, metadata) across cloud and on-premises storage. The system handles encryption at rest and in transit, implements retention policies (automatic deletion after N days/months), and provides compliance-friendly audit logs for data access and deletion.

Provides aggregated analytics and insights across meeting archives, including metrics like meeting frequency, average duration, participant engagement, decision velocity, and action item completion rates. The system generates dashboards and reports that visualize trends over time and enable drill-down into specific meetings or participants.