Matthew Topham
AI Systems & Full-Stack Developer
Boulder, Colorado · thetopham@gmail.com
thetopham.github.io · github.com/thetopham · linkedin.com/in/mrtopham
Computer science student at CU Boulder building full-stack AI systems for interactive storytelling, wearable context, and local inference infrastructure. I work across product interfaces, model integration, persistent state, local compute, and automated validation.
Open-source experience
Marinara Engine + Agents · Pasta Devs
Open-Source Developer · 2026–Present
- Contributed 100 merged pull requests across Marinara Engine (68) and Agents (32), spanning React/TypeScript UI, Node.js services, persistent world systems, agent capabilities, multimodal generation, and automated tests.
- Implemented a working MiniMax H3 / ComfyUI roleplay-animation pipeline: narration becomes a storyboard, character sheets/personas and World Maps artwork ground each scene, generated keyframes anchor the shots, and the final motion plan drives image-to-video generation.
- Live-tested the H3 flow on a successful multi-shot clip with synchronized motion and audio; packaged the proven prompt chain as reusable Storyboard choices.
- Built World Maps/spatial state, reusable character and location references, versioned host APIs for downloadable agents, and recovery paths for stateful workflows.
Selected work: animation pipeline · MiniMax H3 prompts · World Maps · agent capability APIs
Selected projects
Creator · 2026 · Prototype
- Built an Android/Flutter companion and Python services for Omi wearable audio ingestion, daily transcription/lifelogging, searchable notes, and durable upload buffering.
- Connected E09 smart glasses to ChatGPT Live for voice interaction and image uploads through phone automation; voice works today while photo delivery is still being hardened.
Demerzel · Personal AI operations and local fleet
Creator · 2026
- Built tooling for daily AI briefs, LLM wiki ingestion from daily conversations, and management of local AI services.
- Manage an AMD Radeon AI PRO R9700 system, a dual RTX 3060 system, and a single RTX 3060 system for local models, MiniMax H3 video work, and ComfyUI image generation.
- Developed service health checks and workload controls to start/stop supported jobs and avoid GPU contention; separate autonomous coding/research workers remain under refinement.
Built conversational learning experiences with AI historical mentors across web and VR prototypes using React, TypeScript, Node.js, Supabase/Postgres, real-time voice AI, and Unity.
Technical skills
Software: Python, TypeScript, JavaScript, SQL, React, Node.js, Flutter, REST, WebSockets.
AI / infrastructure: local inference, ComfyUI, Linux, Docker, GitHub Actions, Playwright, GPU workload management, regression testing.
Education & certification
University of Colorado Boulder — B.A. Computer Science, expected May 2027
Front Range Community College — A.S. Computer Science, 2025
NVIDIA-Certified Associate — AI Infrastructure & Operations, August 2025
Updated September 6, 2026. Marinara contribution count verified across both repositories September 5, 2026.