Technology

Open where it helps. Engineered for production.

We evaluate tools and models to suit your workload rather than requiring a single proprietary stack.

Technology

Layer	Candidate tools	EXZAI service
Employee AI experience	Open WebUI, purpose-built portals	Authentication, UX, role-specific workflows
Local model runner	Ollama	Local development and prototyping
Production inference	vLLM, SGLang, NVIDIA NIM where licensed	Deployment, benchmarking, batching and scaling
Model gateway	LiteLLM or equivalent	Routing, usage controls, budgets and logging
Retrieval & search	Qdrant, PostgreSQL + pgvector, rerankers	Indexing, retrieval, permissions, evaluation
Workflow & applications	n8n, Python, FastAPI, business APIs	Agents, integrations, human-in-the-loop flows
Packaging & orchestration	Docker, Kubernetes, virtualization	Reliable environments, rollouts, lifecycle
Infrastructure	GPU systems, compute, network, storage	Architecture, procurement, installation
Observability	Prometheus, Grafana, OpenTelemetry	Monitoring, diagnostics, capacity planning
Security & identity	OIDC/SSO, access policies, secrets management, encryption	Secure architecture and operational controls

Technology names indicate tools that may be used in project architectures. They do not indicate commercial partnerships or endorsement.

Illustrative architecture

Model and software licenses are reviewed for commercial use before deployment.
Model endpoints sit behind a reverse proxy, network controls, authentication and authorization.
Document handling, embeddings and telemetry destinations are confirmed before describing a system as fully local.
Citations help verification but do not ensure correctness; evaluation and human review apply where outcomes matter.
Agents do not take destructive or privileged production actions without human approval.

Tell us what you want to build. We will help define a practical technical and commercial path.