# MindSpeak AI
> Building open hardware and local compute architectures for personal AI companions, intelligent agents, and personal memory second brains.

MindSpeak AI (https://mindspeakai.org) creates open, user-owned physical hardware and software systems designed to bring private AI companions into the physical world. By combining acoustic arrays, dedicated audio DSPs, on-device compute, and local LLMs, MindSpeak AI eliminates closed corporate silos, cloud subscription traps, and privacy intrusions.

## Core Offerings

- [DIY Dev Kit](https://mindspeakai.org/devkit): Open hardware PCB dev kit powered by ESP32-S3 and XMOS XVF3800 audio DSP with modular compute slot for private voice assistants.
- [The Toy](https://mindspeakai.org/toy): Child-safe embodied AI companion for interactive storytelling, open-ended curiosity, and adaptive learning without screens.
- [The Tower](https://mindspeakai.org/tower): Ambient home guardian providing private monitoring for elders, infants, and pets with on-device computer vision and voice alerts.
- [The Teacher](https://mindspeakai.org/teacher): Multi-generational conversational knowledge assistant bridging the learning gap for students and seniors.
- [Ollo](https://mindspeakai.org/#products): Second Brain software companion indexing personal memories, notes, and habits locally into a private knowledge graph.
- [AI Bootcamp](https://mindspeakai.org/bootcamp): Intensive 5-day hands-on engineering program teaching developers how to build intelligent agents on existing foundational models and deploy them to physical hardware.
- [Courses](https://mindspeakai.org/course): Curriculums covering Generative AI, MLOps, Computer Vision, and Autonomous Physical Agents.
- [Consumer Labs](https://mindspeakai.org/consumer-labs): Experimental prototypes including AIR, an edge AI thermal-comfort fan that tracks body heat and moves airflow dynamically.
- [Security Labs](https://mindspeakai.org/security-labs): Edge-computing optical intelligence systems like VIGIL, moving beyond passive surveillance into proactive environmental understanding.

## Hardware Architecture

- Baseboard: Custom MindSpeak open architecture
- Microcontroller: Espressif ESP32-S3 (Wi-Fi 4 + Bluetooth 5 LE)
- Audio DSP: XMOS XVF3800 voice processor with 4-microphone far-field circular MEMS acoustic array
- Audio Algorithms: Acoustic Echo Cancellation (AEC), beamforming, dereverberation, noise suppression
- Compute Module Slot: High-density board-to-board connector compatible with Raspberry Pi CM4/CM5, Radxa CM3, or NVIDIA Jetson modules
- Power & I/O: USB Type-C power delivery, stereo DAC audio output, I2C/SPI expansion headers

## Privacy & Local AI Philosophy

- Privacy-First: All voice processing, wake-word detection, and memory indexing can run 100% locally on-device.
- Zero Cloud Lock-in: Compatible with open-source LLMs (Llama 3, Mistral, Gemma, Phi-3, Qwen) running via Ollama, llama.cpp, or local vLLM instances.
- Data Sovereignty: User memories, audio recordings, and knowledge graphs remain stored on local encrypted storage owned by the user.

## Organization & Leadership

- Founded: Originally conceived in 2016; relaunched in 2025 by Founder Chaal Pritam.
- Headquarters: Bangalore / Chennai, India.
- Website: https://mindspeakai.org
- Twitter / X: https://twitter.com/mindspeakai
- Founder Twitter / X: https://twitter.com/chaalpritam
- GitHub: https://github.com/mindspeakai