Building knowledge assistants and practical AI systems.
I’m Michael Lvov. My current project is a knowledge assistant built around retrieval, tool use, and durable conversation state.
On this page
Current work
A knowledge assistant.
The project explores how an assistant can work with prepared knowledge, use tools, and retain conversation state. Its public harness makes the architecture and checks available to examine.
How the knowledge assistant fits together.
A compact view of the published Harness architecture. Browser requests enter directly; MCP clients use the gateway. Manual ingestion prepares knowledge outside the conversation.
Explore the full architectureResearch case study
AI Harness Landscape
Compares how AI products connect knowledge, tools, durable state, and human review into useful work.
Read the case studyStart with a concrete use case to guide what the system needs to do.
Keep responsibilities clear. Make each responsibility explicit so the system is easier to understand and change.
Make the work inspectable. Keep source, architectural decisions, and checks available for review.
Founders
The person behind the lab.
I lead the knowledge-assistant project at Nora SuperIntelligence Lab, exploring how people, knowledge, agents, and tools can work together without losing human judgment.
Contact
Discuss a role or project.
I’m open to technical leadership roles and AI or data-platform projects. The Founders page has the fuller profile and contact context.