About
Hallo, ich bin Ralph. I'm a recent physics grad of Ateneo and will soon begin grad studies. Currently, I work part-time as a consulting RA at a research group in GUST. My work there involves investigating fdtd and terahertz spectroscopy, as well as leading the group's ML efforts.
My interests lie in theoretical physics and technical AI safety (mech interp, ML-physics correspondence). I also have extensive experience in experimental particle, computational, and mathematical physics. My undergrad thesis focused on developing methods for solving nonlinear partial differential equations.
now
Previous 2025-03-15
Updated 2025-07-07
Inspired by Sivers. A snippet of what's up with me lately.
It is 2025 W28 and rainy af at Quezon City, PH. I am currently
- listening to boy pablo (2020), Wachito Rico (a bop, no?),
- investigating diesel oil and terahertz spectroscopy,
- writing a paper on partial differential equations,
- preparing to relocate, and
- upskilling in technical AI safety.
Previously, I was listening to WHO CARES?, presenting a poster and paper at SPP, and wrapping up my bachelors.
next
Updated 2025-07-07
Inspired by Gross. A list of ideas that I'd love to see happen.
- More studies on ML-physics correspondence for AI safety
- A cli-based voice-enabled AI assistant tool a la Gemini Live
- A cloud-based private alternative to Google Sheets
- with more research ideas to come as I read...
collab
Inspired by Lenz and Perez. A breakdown of my working flow and style.
Email me if you wanna work on something.
On working alone, I typically
- work in the terminal on my arch-based macbook,
- use plain text and simpler formats as much as possible (eg. markdown for notes over gdocs, typst for papers and slides over latex, marimo for experiments over jupyter), and
- work in large time blocks each day.
On working with collaborators, I'd want us to
- give honest and prompt feedback on anything that could be improved,
- meet 1-3 times a week or less and work primarily asynchronously,
- use cli- or cloud-based tools (gsuite is fine), esp over those that aren't linux-friendly, and
- align often (in depth, not necessarily agreement) and with full context before execution.
On working with mentees, I prefer that you
- over-communicate how you feel about your progress and the project,
- take ownership and organize whatever you think is helpful for you (eg. coworking, reading, discussions), and
- build a strong mental model of what your mentor would say (eg. regarding next steps on the project, feedback on results).
links
Inspired by no one lol. A collection of quick links for your convenience.
And some recos