🧠 Founder, MemorySync · Memory infrastructure for autonomous systems

Agents don't fail because the model is weak. They fail because their state is mismanaged. I build the layer that fixes that.

Context windows are finite and tokens are a budget. I work on keeping long-running agents fast, cheap, and coherent:

• 🗄️ Long-term retention that outlives the context window and carries across sessions and tools
• 🕸️ Memory graphs that retrieve by relationship and meaning, not just vector similarity
• 🎯 Adaptive retrieval that returns the smallest useful slice instead of everything that loosely matches
• 🔥 Hot / warm / cold tiering that moves memory between tiers based on how it is actually used, so token bloat stays under control

MemorySync is production memory infrastructure for AI agents.

• 🔒 Core engine: a managed service
• 🔌 Framework adapters: fully open-source at @memorysyncio, with native support for Claude Code, Cursor, n8n, Dify, Pipecat, LangGraph, Agno, Mastra, Vercel AI SDK, and CrewAI

Vector index runtimes for low-latency recall, async streaming pipelines that keep memory writes off the agent's hot path, and containerized server architectures built for reproducible deployment.