01 / The premise
How can an agent retain useful context without making its behavior opaque?
Development and integration of AI-powered agentic systems, including vector embeddings, persistent memory, MCP clients, and automated code-generation workflows deployed on Azure.
02 / What took shape
- Semantic retrieval pipelines with ChromaDB
- Persistent memory for long-running agents
- MCP integrations and Azure deployment
03 / Looking back
The useful residue.
Agent quality depends as much on context boundaries and observable state as it does on model capability. Memory is a product surface, not merely infrastructure.