AI agents
Tool-enabled agents for multi-step workflows and controlled autonomy.
Agents, local models and hybrid AI architectures for real products and business workflows — with privacy built into the system design.
Discuss an AI projectPrivate zone
Local models
Private knowledge
Capability zone
Frontier models
Controlled access
Policy & routing
Privacy · Capability · Cost
Applied AI
Tool-enabled agents for multi-step workflows and controlled autonomy.
Cloud and local models inside products, tools and business processes.
APIs, tools, scheduled workflows and connected systems.
Retrieval, embeddings, context engineering and persistent memory.
Privacy-first AI
For German and European companies, model quality is only one part of the architecture. I also design for data sovereignty, provider independence and deliberate access control.
Hybrid AI
Hybrid routing lets each task use the right path instead of forcing one model onto every workload.
Built in practice
My own product for working with multiple models, specialised agents, tools and local/cloud execution in one coherent system.
Explore GraceWorking stack
Models
OpenAI · Claude · Gemini · DeepSeek · Qwen · Llama
Agents
Tool Calling · MCP · Multi-Agent · Structured Outputs
Knowledge
RAG · Embeddings · Vector Search · Context Engineering
Infrastructure
Local inference · Self-hosting · Docker · Hybrid setups
Engineering
TypeScript · Python · React · Next.js · Supabase