I research and engineer language models, latent reasoning architectures and local AI systems, then build the software that turns those ideas into usable products.
Recent work spans coding models, privacy-first agent infrastructure, API compatibility layers, realtime systems and GPU-aware desktop software. The newest projects lead; the archive keeps the rest one click away.
A coding-focused 9B model released for local inference, with GGUF quantizations ranging from Q8 to extreme low-bit builds. The release includes an importance-matrix workflow tuned on code and conversation data, plus a dedicated product landing page.
Open model ↗ Open landing page ↗A privacy-first Windows runtime that separates model reasoning from host authority using Hyper-V isolation, deterministic policy, audited tools and VPN-gated networking.
A local bridge that exposes OpenAI Chat Completions, OpenAI Responses and Anthropic Messages over one AutoClaw-backed setup, with streaming, tools and client-specific integration helpers.
A unified Windows launcher for compatible PS4/PS5 content with modular cores, PKG import, Vulkan tuning, hardware diagnostics and frame-time-focused AutoBoost policies.
An open-source realtime community app built with Flask, SQLite, Socket.IO and WebRTC, covering servers, DMs, forums, calls, screen sharing, bots and webhooks.
Recurrent latent computation for frozen decoder models through a structured J-Space adapter, competing branches, verification and adaptive halting.
An AI-powered educational ecosystem connecting students, teachers and families through intelligent tools, dashboards, media, gamification and school workflows.
Local AI music generation, songwriting assistance and GPU-aware model orchestration in one workstation-first environment.
A modular desktop browser around QtWebEngine with isolated profiles, persistent sessions, local data and custom browser UX.
A local knowledge assistant combining semantic retrieval, BM25/FTS, rank fusion and traceable source references.
Compact coding model fine-tuned for debugging, structured reasoning, tool use and local agentic workflows.
Gemma-based coding model with visual debugging support and GGUF-oriented local inference workflows.
The portfolio intentionally crosses research and product engineering. Understanding the model matters. Understanding the runtime, interface and constraints around it matters too.
Fine-tuning, coding models, evaluation, quantization, GGUF publishing and practical local inference workflows.
Sandboxed execution, capability-scoped tools, deterministic policy, audit trails and provider-agnostic orchestration around probabilistic models.
Applications and compatibility layers that keep inference, credentials and workflows close to the workstation.
Structured internal representations, recurrent depth, adaptive compute and experiments that modify how models deliberate before decoding.
Socket.IO, WebRTC, browser media, local APIs and desktop/web systems with live state and multi-user interaction.
RAG systems combining semantic search, lexical retrieval, reranking and traceable sources around local or hosted models.
Systems connecting language, vision, audio and 3D generation instead of treating each model as an isolated demo.
Desktop tools, browsers, emulation frontends, dashboards and end-to-end products built around real runtime constraints.
My current research direction explores whether structured recurrent computation inside a constrained latent subspace can improve a frozen decoder under controlled paired evaluation.
LeBRON adds a lightweight recurrent reasoner around a frozen decoder. A sparse J-Space projection exposes a structured workspace, competing branches evolve internally, and adaptive halting decides when additional compute stops being useful.
I care more about choosing the right abstraction than collecting logos, but these are the tools that repeatedly survive contact with my projects.
I'm an AI Engineer and researcher working across coding language models, latent reasoning, local AI, agent infrastructure, multimodal systems and end-to-end software.
I build both the models and the systems around them, moving between research experiments and practical engineering.
My work explores what happens inside intelligent systems, how their internal representations can be structured, and how those ideas can become useful tools in the real world.