Computer Science graduate (AI specialisation), Baguio City, Philippines.
From April 2025 to August 2026, I was Project Assistant VI on Project CrimeXPerience, a DOST-funded XR crime scene simulation at the University of the Cordilleras. I built Gemini Live Audio witness interviews with rubric grading and scripted fallback dialogue, a server-side token broker, and access gates requiring operator and instructor authorization. I also worked on Unity AR and VR simulation features, Firebase integration and Python data processing, and trained staff to run the simulations. The project's code is private.
- SQL Query Agent: Claude tool calling, a Cloudflare Worker and D1 demo data, with a keyword-blocklist SQL guard. Source and guard tests are available; there is no public deployment.
- Job posting intake: an n8n workflow with Gemini extraction, field validation and Google Sheets output. The fetch path has runtime evidence; the model/Sheets path is not yet verified. Make and Zapier versions are planned.
- Gameplay video QA: Python/FFmpeg measurements, HUD-bar readings and labelled evidence for reviewing recorded gameplay. Synthetic-video tests check the tools; a person reviews bug findings.
- Operations dashboard: versioned CSV imports, maintenance assignments and finance/leasing summaries. A local reference implementation with fictional data and local reminder logs.
- Lead-to-CRM automation: source-backed assessment, human approval and a durable operation queue. The offline demo uses fixtures; connected HubSpot/Instantly execution is unverified.
- Rust calculator: a desktop calculator with mouse/keyboard input, expression parsing and formatting tests.
- Late-trip monitor: a check for medical-transport dispatch that compares driver ETAs with pickup windows and drafts rider texts for a dispatcher to approve. It uses made-up data and has an interactive demo; it isn't connected to a dispatch system.
- Job tracker: a Rust/egui desktop app and command line over an Excel application log, with save checks, backups and tests that render each screen.
I have been modding Total War since 2020, using Python generators, relational game data and Lua campaign scripts. Projects include The Grand Trade Exchange, Chaos Dwarf House Ancillaries, The Great Guilds and The Iron Court.
The mods page links the other projects and separates out-of-game checks from in-game testing. Each repository records its own scope, checks and Workshop status.
The video showcase contains AI-generated spec work for fictional brands, with production notes. It is not client work.
Programmatic video editing shows three narrated explainers made without AI video: openly licensed footage, Manim graphics, and cuts placed in code by the narrator's word timings. The narration is Gemini text-to-speech. The Great Pyramid video also has a second cut built in Remotion (source).
From January to August 2024, I led research on photogrammetry and Neural Radiance Fields for cultural preservation. I built Python image-processing pipelines and benchmarked Instant-NGP, Splatfacto and Nerfacto against photogrammetry using STD, RMSE and MAE. Presented at PCSC and AUDRN.

