Blog

I was building an agentic personal assistant and hit the wall everyone hits - my agents kept forgetting everything. Context lost mid-conversation. Zero recall across sessions. No ability to connect the dots between people, projects, and deadlines. So I built a 3-layer memory architecture inspired by how human cognition actually works - conversation context with smart compaction, short-term memory that persists across sessions, and a long-term knowledge graph where edges decay and strengthen like real synapses.

How I deployed Small Language Models, built agentic workflows with LangGraph, and operationalized vLLM inference engines at PKSI to deliver scalable, cost-efficient enterprise AI solutions.

A deep dive into my internship at Van Brunt & Associates, where I built ETL pipelines with Airflow, developed load forecasting models for the ERCOT grid, and identified critical 4CP events to drive cost savings in the Texas energy market.

Exploring network visualization with Neo4j, Gephi and microsoft GraphRAG, enriched by LLM-driven semantic analysis. A comparison of tools and techniques for graph-based knowledge representation and storytelling.