AI Engineer · Systems Architect
I build autonomous AI agents that ship — from a Blender 5.0 operator to a voice-first VS Code pair programmer. Official contributor to Google Gemini CLI and Vercel AI SDK.
Artificial Intelligence systems engineer focused on autonomous agents and cost-efficient orchestration. I specialize in tool-grounding, retrieval-augmented generation, and multi-turn agentic workflows — the plumbing that turns LLMs into reliable, shippable systems.
Today I'm building Sequence 3D (autonomous 3D modeling agent), Sequence Code (autonomous coding agent), Sequence RPG (multi-agent AI RPG), and Sequence Live (real-time live assistant). Along the way I've shipped official contributions to Google's Gemini CLI and Vercel's AI SDK.
Source of truth for every project on this site — Sequence 3D, Sequence Code, Sequentia, SNIP-DIFF, and this portfolio.
Deterministic, stateless desktop app for intelligent code diffing and large-file summarization — real-time syntax highlighting and diff analysis without the multi-turn overhead of an LLM chat loop.
Write-ups on agent architecture, PR walkthroughs, and demo videos of the shipped Gemini CLI fixes.
AI operator add-on for Blender 5.0. Drives Blender through a deterministic tool layer — primitives, transforms, selection, mesh edits, modifiers, materials, physics. Currently perfecting tool grounding and the ability to author its own Blender operators in Python on the fly. Powered by Gemini 3.0 Pro with a 2M context window.
Voice-first AI pair programmer that lives inside VS Code. Remembers every session, adapts to your learning style, and pairs with you through all 14 core algorithmic patterns — spoken, not typed. A coding agent that teaches instead of just completing.
AI-driven RPG where every NPC is a persistent LLM agent with its own goals, memory, and evolving relationships. The agent layer is feature-complete — NPCs perceive nearby entities and overhear conversations within a radius. Currently deep into world and asset building: physics, villages, map, and shaders.
Real-time live assistant for conversations. Understands what you're looking at and answers in real time. Dual-agent architecture — a quick first response lands fast, followed by a deeper follow-up. Dual-channel audio in and out, and an intuitive HUD that stays out of your way. Everything works end-to-end; scaling, authentication, and launch are next.
Hobbyist 3D modeling and digital art. Blender is my sketchpad — the same tool Sequence 3D targets. Studying topology, materials, and lighting the long way so my agent work stays grounded in the real craft.
Tinkering with engines, scripts, and asset pipelines. Good practice for reverse-engineering systems and respecting the seams between game logic and content.
What I'm actually sinking hours into.