RM
Ralph Morales
~/portfolio/projects

hello, world

hey, I'm Ralph

Learning Python, backend systems, and cloud security fundamentals by building.

2nd year BSCS student at Mapua Makati, based in Manila. Right now I am focused on getting better at Python, backend systems, cloud basics, and defensive workflows by building projects and slowly learning how to explain what is going on.

PythonNode.jsFastAPIExpressDockeropen to internships
lightnovel-voiceFeb 2026 to April 2026
● live

78% accuracy in narration/dialogue classification on test set of 10 chapters from 5 different light novels

An AI-assisted light novel reader that separates narration from dialogue and drives more expressive voice output for each speaker.

  • Reached 78% narration vs dialogue classification accuracy on a small held-out test set.
  • Supports voice cloning from a 3-second reference clip.
GPT-4o-miniQwen3-TTSNode.jsExpress
view repository
ftl-multiverse-macApril 2026
● live

Turned a "borderline impossible" native macOS mod path into a reproducible installer and troubleshooting workflow

A reproducible native macOS install workflow for FTL: Multiverse that builds Hyperspace from source, patches a real Mac FTL.app, and documents how to maintain the setup over time.

  • Validated against a real native Mac FTL 1.6.13 app and reached the Multiverse main menu.
  • Built helper scripts for one-shot install, extra mods, pinned version updates, and smoke-test validation.
BashCMakeSDL2FTLMancodesignGitHub Actions-ready workflow
view repository
QuestlineApr 2026 to present
◌ building

Genshin-inspired tracker for Philippine government document workflows

A gamified tracker app for Philippine government document processes, designed to make confusing real-world requirements feel more structured and less overwhelming.

  • Grounded in a specific use case: Philippine government document workflows.
  • Uses Supabase, Google Places, and GitHub Models rather than being just a static concept.
FlutterDartTypeScriptExpressSupabaseGitHub Models
view repository
Akashic — AI GalleryFebruary 2026 to March 2026
● live

~120ms avg search latency

Semantic image search for personal photo libraries using natural language instead of folders or filenames.

  • ~120ms average search latency for retrieval.
  • Uses text-embedding-3-small plus GPT-4o-mini for semantic understanding.
GPT-4o-minitext-embedding-3-smallExpressNode.jsExpoReact Native
view repository