I'm a CS Masters' graduate specializing in machine learning and signal processing, with hands-on experience in audio and visual production. Outside of technical work, I founded and run RausHaus Productions.
GPA: 3.67
Specialization: Interactive Multimedia
GPA: 3.74 (Magna Cum Laude)
is a Manila-based media collective comprised of young, passionate, and hardworking artists, photographers, and creatives.
It started as a sort of joke when friends back in high school would consistently rehearse in my bedroom (Rau's House).
When the pandemic hit and everyone was stuck at home, I started to learn how to produce and record music from my bedroom.
What started as a joke eventually became a real studio as I found myself increasingly invested in recording and making music.
Now, RausHaus is an actual recording, mixing, and mastering studio that produces music from demos into actual ready-to-release works.
The Haus is the DJ event arm by Juan Aruego (WEGS) and Gabe Ocampo (GABRIEL). It's partnered with multiple local bars.
Live at RausHaus is our live session format, giving more emphasis on local artists and their original music.
make a core part of RausHaus in championing local Filipino music and giving avenues for our friends hoping to break into the industry through gigs and event launches.
The study presents a modification to the foundation of a noise gate plug-in to create a more intelligent, context-aware system focused on target-speaker speech identification. A supervised, text-independent, and frame-wise personal voice activity detection model based on Ding et al. (2019) is proposed and refined to process audio in real-time with minimal CPU utilization.
The model achieved 87.6% accuracy during batch testing and 72% accuracy on a 5-minute real-world sample under challenging conditions, outperforming a standard noise gate on the same input. Overall, this work demonstrates the feasibility of applying deep learning models in real-time audio systems and demonstrates the potential of personal voice activity detection as a control mechanism for adaptive audio effects, contributing insights to the interdisciplinary fields of computer science and live audio broadcasting.
Technologies: Python, PyTorch, Torch, Google Colab, C++, JUCE Framework
Conference Acceptance: 2026 Joint International Conference on Digital Arts, Media and Technology with ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunication Engineering (ECTI DAMT & NCON)
This paper compares K-Means, DBSCAN, and Affinity Propagation clustering on Queen's discography using full and reduced musical feature sets extracted from MIDI data. Each technique-feature set combination was evaluated with Silhouette Score, Davies-Bouldin Index, and the Elbow Method.
Results show the reduced feature set produced more clusters with fewer outliers, and Affinity Propagation generated more analyzable clusters than K-Means, while DBSCAN underperformed due to the small dataset size.
Overall, the reduced feature set paired with Affinity Propagation gave the most analytically valuable results, with qualitative analysis pointing to areas for improvement such as non-deviating clusters and dataset misrepresentation.
Technologies: Python, scikit-learn, jSymbolic, Google Sheets
This study compares Windows 11, macOS, and Linux Mint as host operating systems for a Type 2 hypervisor, using a triple-boot setup on a single machine. Results show the impact of each host OS on virtual machine performance, thermal behavior, and resource utilization.
Findings highlight trade-offs between raw performance, thermal efficiency, and resource allocation across the three systems, offering guidance and trade-offs on selecting an optimal host OS for virtualization.
Technologies: VirtualBox, Boot Camp, Geekbench, PassMark, RAMSpeed, MPrime
This study examines player popularity and fan discourse surrounding Gilas Pilipinas on YouTube and Rappler from July to September 2024. Using web scraping and sentiment analysis, the study identifies the most discussed players, sentiment trends per player, frequently associated keywords and language patterns, and key differences in tone and discourse style between the two platforms.
Findings highlight how platform and audience shape sports discourse, offering insight into the cultural dynamics of Gilas Pilipinas fandom in the Philippines.
Technologies: Python, Beautiful Soup, YouTube API, NLTK, scikit-learn
Built privately for RausHaus Productions, this pipeline automates the creation of karaoke-style subtitles for music video releases.
Isolated vocal audio and corresponding lyrics are fed into stable-ts, a Whisper-based forced alignment tool, to generate frame-accurate timing for each line and word. The aligned output is converted to .ass subtitle format, manually verified for accuracy in Aegisub, and finally converted to .ytt for direct upload to YouTube.
This tool has been used and customized privately for RausHaus Productions' kareoke-style video releases, streamlining a process that would otherwise require manual subtitle timing for every release.
Technologies: Python, stable-ts, Aegisub, YTSubConverter
Audio processing, feature extraction, speaker recognition, machine learning, and neural networks. Python and PyTorch for machine learning.
HTML, CSS, JavaScript fundamentals, Git, command line basics, and responsible problem-solving for beginning web development.
Ruby fundamentals, Rails framework, databases, full-stack web application development, and deployment.
LinkedIn
Google Docs
Google Drive
Google Forms
Google Sheets
Google Slides
OneDrive
Zapier
Zoom
Google Apps Script
C++
CSS
HTML
Java
JavaScript
Python
SQL
Premiere Pro
After Effects
DaVinci Resolve
Final Cut Pro
Photoshop
Illustrator
Canva
Figma
OBS Studio
Ableton Live
FL Studio
Logic Pro
Max 8
Reaper
Wwise