Edge AI systems
Applied AI designed for inference close to the operating environment, with real constraints on hardware, latency, power, and connectivity.
Applied AI program
AI LAB
Overview
From constrained inference to IDE-native agent tooling.
AI Lab focuses on practical AI systems that can be deployed close to the operating environment, optimized for constrained hardware, and integrated into real engineering workflows.
Quantization and optimization
Model quantization, compression, and runtime optimization to move useful capabilities onto smaller devices without losing operational control.
Developer tooling for agent workflows
IDE-native tooling that gives coding agents and engineering teams durable context, code-linked memory, and structured retrieval across sessions.
Shipped tooling
NeuroTrace MCP
Persistent memory for coding agents in IDEs.
NeuroTrace MCP is a local-first, code-linked memory layer that helps agents and developers preserve project context across chats, sessions, and files.
It is built as developer tooling for agent workflows: structured memory, retrieval before coding, and workspace-linked context that stays with the repository instead of the chat.
Stores high-signal decisions, tasks, and project context across chats and sessions.
Keeps memory attached to the repository, files, and active workspace instead of the chat alone.
Generates workspace guidance and host-specific configuration for supported IDE agent workflows.
Generated setup
- Local .neurotrace workspace directory and memory database
- MCP templates and host-specific setup guidance
- Workspace-linked memory flows for long-running codebases