Senior Software Engineer · Builder · Medellín, Colombia 🇨🇴

10+ years building scalable web applications and digital products. I care about product engineering, shipping real value, and leveraging AI to solve actual human problems. Unit economics, robust architecture, and great UX matter.

Core  

AI & Data  

Infrastructure  

A multi-thread, persona-based conversational interface with Deep Memory. Instead of traditional RAG, it uses a Temporal Knowledge Graph (Neo4j + Graphiti) and a custom Context Hydration pipeline to maintain persistent psychological context across infinite conversation threads.

React 19 · Convex · Python FastAPI · Neo4j · Graphiti

An AI-powered nutrition tracker that owns its raw data. It understands text and meal photos, logs macros into Supabase, and syncs with Google Sheets — no subscription, no walled garden.

Python · LangGraph · Telegram · Supabase · Google Sheets API

I write about the engineering reality of building AI products — the tradeoffs, the architecture decisions, and what actually works.

• My AI Sends 30k Tokens Per Message. 80% of Them Were Wasted. — 👁️ 30
• My Wife Sent 297 Messages in 15 Days. Not to Me. To the AI I Built Her. The Synapse Story — 👁️ 64
• I Benchmarked Graphiti vs Mem0: The Hidden Cost of Context Blindness in AI Memory — 👁️ 216 · 12 min read
• Full Circle: Giving My AI's Knowledge Graph a Notion Interface Using MCP — 👁️ 1,621 · 5 min read
• Scaling AI Memory: How I Tamed a 120k-Token Prompt with Deterministic GraphRAG — 👁️ 356 · 5 min read
• When 5 Minutes Isn't Enough: Moving AI Ingestion from Sync to Async — 👁️ 50 · 4 min read
• Beyond RAG: Building an AI Companion with Deep Memory Using Knowledge Graphs — 👁️ 1,514 · 7 min read
• I Used My AI Nutrition Agent Every Day for a Month. Here's What I Actually Had to Fix — 👁️ 73 · 6 min read
• I Ditched MyFitnessPal and Built an AI Agent to Track My Food — 👁️ 247 · 8 min read

→ All articles on Dev.to