ScaiInfer — Inference Compute

ScaiInfer provides the GPU-powered compute infrastructure that serves AI model inference. ScaiInfer nodes register with ScaiGrid and handle the actual model execution — loading models into GPU memory, processing requests, and returning results.

What makes ScaiInfer flexible is where it runs. Nodes can operate on ScaiLabs' own infrastructure, on a partner's data center hardware, or on-premises at a customer site. ScaiGrid handles the routing transparently — applications don't know or care which node processes their request.

Key capabilities:

GPU-optimized inference serving
Automatic node registration with ScaiGrid
Health monitoring and capacity reporting
Support for multiple concurrent models per node
Hardware flexibility: cloud, data center, or on-premises

ScaiAtlas — Model Registry

ScaiAtlas is the central catalog for all AI models in the ScaiLabs ecosystem. When a model is trained via ScaiMind, imported from an external source, or downloaded from a model hub, it's registered in ScaiAtlas with full versioning metadata.

ScaiAtlas manages model lifecycle: storage, versioning, discovery, compatibility metadata, and deployment coordination to ScaiInfer nodes. It ensures that everyone — from developers to ScaiGrid's routing engine — knows exactly which models are available, what version they are, and where they're deployed.

Key capabilities:

Model storage and versioning
Metadata management (architecture, capabilities, hardware requirements)
Deployment coordination with ScaiInfer
Model discovery and search
Access control integrated with ScaiKey

ScaiMatrix — Vector Store & Semantic Search

ScaiMatrix provides multi-tenant vector storage for embeddings-based search and retrieval-augmented generation (RAG). It handles the full pipeline from raw documents to searchable knowledge.

When a document enters ScaiMatrix, it's processed through a configurable pipeline: chunking (splitting into meaningful segments), embedding (generating vector representations via ScaiGrid), and indexing. The result is a searchable knowledge base that powers RAG for ScaiBot, ScaiCore agents, ScaiWave AI participants, and any application connected to ScaiGrid.

Key capabilities:

Multi-tenant vector storage with tenant isolation
Document ingestion pipeline: chunking, embedding, indexing
Semantic search with configurable similarity thresholds
Integration with ScaiGrid for embedding generation
Collections management: create, update, delete knowledge bases
Security trimming: search results respect ScaiKey permissions
API and web interface for knowledge management

ScaiMind — Training & Fine-Tuning

ScaiMind orchestrates model training and fine-tuning workloads across GPU-powered infrastructure. Where ScaiInfer handles inference (using trained models), ScaiMind handles the process of creating and improving those models.

ScaiMind supports multiple training approaches — from lightweight LoRA adapters to full fine-tuning — and manages the entire lifecycle: dataset preparation, training job scheduling, GPU resource allocation, progress monitoring, evaluation, and model registration in ScaiAtlas.

Key capabilities:

Training job orchestration across multi-GPU nodes
LoRA, QLoRA, and full fine-tuning support
Dataset management with ScaiDrive integration
Training monitoring: loss curves, evaluation metrics, resource utilization
Automatic model registration in ScaiAtlas upon completion
Evaluation and benchmarking against reference datasets
ScaiGrid module interface for management and monitoring

ScaiVault — Secrets Management

ScaiVault provides secure storage and lifecycle management for secrets, credentials, and certificates across the ScaiLabs ecosystem. Every service that needs API keys, database credentials, encryption keys, or TLS certificates retrieves them from ScaiVault.

ScaiVault supports both manual secret management and automated workflows: credential rotation on schedule, PKI certificate issuance and renewal, and agent-based secret injection that delivers credentials to services without exposing them in configuration files.

Key capabilities:

Encrypted secret storage with access control via ScaiKey
Automated credential rotation
PKI capabilities: certificate issuance and management
Agent-based secret injection for services
Audit trail for all secret access
API-first with management UI

ScaiSend — Notification Delivery

ScaiSend handles transactional email and notification delivery for the entire ScaiLabs ecosystem. It replaces external services like SendGrid with an infrastructure-native solution that integrates with ScaiKey for identity and ScaiVault for credentials.

Every notification in the platform — password resets, approval requests, escalation alerts, system notifications, webhook deliveries — flows through ScaiSend.

Key capabilities:

Transactional email delivery with template management
Multi-channel notifications (email, webhook, SMS)
Delivery tracking and retry logic
Template engine with variable substitution
Integration with ScaiKey for sender identity
API for programmatic notification from any service

ScaiDNS — DNS Management

ScaiDNS provides multi-tenant DNS zone management built on PowerDNS. It's a management layer — not a replacement for DNS infrastructure itself. ScaiDNS manages configuration, access control, and provisioning; DNS queries flow through PowerDNS directly.

ScaiDNS supports forward and reverse zones (IPv4 and IPv6), DNSSEC, domain ownership validation, and fine-grained access control at both domain and record level. It stores DNSSEC key material internally (rather than in ScaiVault) to avoid bootstrap dependencies during infrastructure initialization.

Key capabilities:

Forward and reverse zone management (IPv4 and IPv6)
DNSSEC support with internal key storage
Domain ownership validation via DNS (no WHOIS dependency)
Multi-tenant isolation via ScaiKey
Domain-level and record-level access control
All record types supported by PowerDNS
API with API key support for external integrations
MCP server for AI-native DNS management

Learn more

View Documentation →