portfolio
LinkedIn | GitHub | Analytics blog (Medium) | Rotating fish
hello! im riki!
I have a Master’s in analytics from Georgia Tech and am a former IMF research analyst. I’ve spent 5+ years as a full-stack research and data person across economics and international policy institutions. Currently, I lead an ML project that predicts farmers’ adoption rates of agricultural training deployed for project teams.
I try to continuously upskill in ML/AI tooling. For a project, I used LangChain + ChromaDB to build an LLM-based app that assessed foreign investment suitability. Most recently, I wrote an essay about how camera resolution and distance affect facial recognition accuracy.
Outside my career, I work on my art projects, practice judo, volunteer in the DC Community Emergency Response Team (CERT), and am prepping for my amateur HAM radio technician license.
what I’m thinking about right now:
Point metrics don’t tell you whether the model outputs align with the observed data distribution.
I.e., if the model has a good (low) RMSE but predicts a noticeably different distribution, it might be overfitting the center and neglecting the tails. Distributional metrics help identify this.
[image or embed] — riki (@rikimatsumoto.bsky.social) September 5, 2025 at 1:47 PM