Synthetic Intelligence Interactive Matrix Personal Adaptive Familiar
SIIMPAF is a comprehensive self-hosted AI infrastructure that provides document processing, vector-based search, and intelligent content analysis capabilities using consumer hardware and open-source tools. The name bridges modern technology with the mystical concept of a familiar - a helpful companion that assists in daily tasks.
The system consists of a Python backend with FastAPI REST API server and integrates with tools like Ollama, vLLM, and Qdrant for private, self-hosted AI resources.
Process and analyze documents with AI-powered extraction, summarization, and semantic search.
Retrieval-augmented generation using Qdrant vector database for context-aware responses.
Animated AI avatars using SadTalker, EMAGE, and PantoMatrix for lip-sync and body motion.
Supports Ollama, vLLM, and LoRA fine-tuned models for different use cases.
High-quality voice synthesis with Coqui TTS, Edge TTS, and voice cloning support.
GPU workloads distributed across DGPUNET Ray cluster for demanding tasks.
mistral, llama2, dolphin-mistral, nomic-embed-textRPEPTFS demonstrates SIIMPAF's capabilities in a real-world application: an AI-powered pitch training platform where entrepreneurs practice their investor presentations with realistic AI-generated investor NPCs.
Each investor NPC has a distinct personality, investment focus, and questioning style powered by QLoRA fine-tuned LLMs. The system generates animated avatar responses complete with lip-sync, facial expressions, and body gestures - creating an immersive training environment that prepares founders for actual investor meetings.
Why DGPUNET is Essential: Running 5-6 simultaneous AI investors with real-time animation would overwhelm a single GPU. The nv5090's 32GB VRAM can handle only 2 NPCs at once. By distributing the workload across DGPUNET's 92GB of total VRAM, RPEPTFS can deliver the full multi-NPC experience that makes pitch training realistic and valuable.
SIIMPAF's modular architecture enables diverse applications beyond RPEPTFS:
AI tutoring assistants for learning management systems with context-aware responses.
Intelligent practice companions for Practicing Musician's platform.
AI-powered NPCs for therapeutic role-playing game sessions.
RAG-powered research assistants for large document repositories.
The following articles document the journey and philosophy behind SIIMPAF and related AI infrastructure:
October 2025
A philosophical statement about accessibility in AI development - building distributed GPU infrastructure using consumer hardware when cloud providers gatekeep resources.
October 2025 | ~12 min read
Early experiences with pattern matching, NPC behaviors, and making computers feel responsive laid the foundation for four decades of AI development.
October 2025 | ~15 min read
IRC bots for natural language processing, Beowulf clusters, building ISPs and data centers - commodity over enterprise, distributed over centralized.
October 2025
Therapeutic gaming, educational technology, and real-time AI systems that outperformed commercial solutions.
October 2025
GPU scarcity, centralization concerns, and building DGPUNET to democratize AI computational resources.
October 2025
Bringing four decades of lessons together in SIIMPAF - a comprehensive self-hosted AI system embodying distributed computing and computational independence.
October 2025
From IRC bots and NPC behaviors to RPEPTFS - an AI-powered pitch training platform with QLoRA-trained investor NPCs.