Thoughts on AI, systems design, and building intelligent software.
Local LLMs like Llama 3 have a "goldfish memory" problem. Standard RAG helps, but it lacks continuity. Here’s a deep dive into my 4-Tier Cognitive Architecture (Working, Episodic, Semantic, Principles) that gives Ollama intent-aware, infinite context—solving race conditions, circular dependencies, and retrieval fusion along the way.
Financial regulations change daily. Our codebases change hourly. Who keeps them in sync? Introducing Compliance DNA—a graph-based engine that traces the lineage from legal text down to individual function calls.
GitHub solves code storage. But who solves team orchestration? Introducing Nexora, an open-source AI platform built to manage the "wetware" of software development—psychological safety, skill velocity, and team composition.
Why stop at one brain? Exploring the shift from singular LLMs to collaborative multi-agent swarms. We look at orchestration patterns, shared memory, and the "herding cats" problem of AI engineering.
Deep dive into how I built this portfolio using Next.js 15, Supabase, and a custom 3D neural interface. From security hardening to "sketch-style" UI physics.