About | TranscriptIntel

Gedank Rayze

TranscriptIntel is built by Gedank Rayze, a software engineering company based in Lisbon, Portugal. We build tools for teams that work with conversation data at scale.

Our focus is on the infrastructure layer: the runtime, the analysis pipeline, the data model. We believe the best tools disappear into the workflow and let researchers focus on insights, not software.

That work spans the full stack of conversation intelligence — from local-first transcription and document extraction to multi-agent analysis and synthetic data generation — built on a suite of Rust tools engineered for speed, privacy, and reliability. TranscriptIntel is where those pieces come together into a single platform.

Lisbon, Portugal

+351 938 42 53 42

gedankrayze.com

Background

The experience that shaped the platform.

The founding team includes a former Cellebrite engineer with deep expertise in digital forensics and conversation intelligence. That background shapes everything about how TranscriptIntel handles data: chain of custody matters, evidence traceability is not optional, and analysis must be reproducible.

Working in forensics taught us that conversation analysis tools need to work across domains. The same patterns that matter in an interrogation transcript matter in a patient interview or a consumer focus group. The behavioral signals are universal; only the terminology changes.

Vision

Where we are and where we are going.

Domain-Agnostic

No hardcoded industry logic. Behavioral segments, analysis flows, and generation configs adapt to any vertical. Forensics, pharma, market research, HR, legal, education.

28 Agent Crews Live, 42 Total

28 agent crews are live in production and 14 more are in progress. Each crew is purpose-built for a specific analysis task.

Building Honestly

We show what is implemented and what is in progress. No vaporware. The agent-crew roadmap on our Features page has real status badges, not marketing promises.

Technology

Purpose-built tools for conversation intelligence.

IronCrew

Multi-agent orchestration runtime written in Rust. Stateless compute with SSE streaming, per-run abort, model routing, and prompt caching. Coordinates multiple LLM calls into structured analysis flows with error recovery.

IronFlow

DAG-based workflow automation engine in Rust with Lua scripting. 96 built-in nodes for HTTP, file I/O, database, AI, and document extraction. Runs complex multi-step pipelines with parallel execution, retry logic, and conditional routing. Ships as a single binary.

transcribeit

Rust CLI for local-first transcription with speaker diarization and VAD segmentation. Supports whisper.cpp, sherpa-onnx, OpenAI, and Azure providers. Audio never leaves your machine. Any audio/video format via FFmpeg.

Cognitive-OCR

Cognition-aware document extraction engine in Rust. Transforms PDF, DOCX, PPTX, and images into structured Markdown using local ONNX models for layout detection and vision-language models for semantic understanding. Triages documents first, then applies the right extraction strategy per page.

Built by Engineers Who Understand the Problem