Reza Ramji

Hi! I'm Reza, and I'm interested in ancient history, math, historical monetary policy, numismatics, and machine learning; currently an undergrad @ princeton classics

Reach me at rezaramji@princeton.edu or on LinkedIn.

Projects

Frontier lab work

· link

Contracted as a consultant by a VC-backed startup to work on and assist proprietary autoformalization techniques.

Euclid autoformalization

read · PDF

I tested how different translations and promptings impact an AI/LLMs ability to take Euclid's geometric proofs(book I) into Lean 4 (A checkable math programming language). I found the prompt's example style mattered far more than the translation, and learned a lot about testing for contamination and designing rigorous tests.

Youklid

open · source

An AI-annotated reader of Euclid's Elements Book I to find new relationships of ideas in the text; every proposition is annotated with the exact dependencies. The reader also supports Greek and English. I really enjoyed making AIs perform and verify logical steps.

Logion with weighted Levenshtein

read · PDF · source

Princeton's Logion suggests restorations for damaged ancient Greek texts; however, it treats every letter swap as equally likely. I tested an improvement swap based on how easily scribes would actually confuse letters. This improvement performed significantly better in a short test on Photius.

Startup name generator

open

Retrained karpathy's microgpt on startup names

llm-eval

source

A small A/B harness for comparing language models across providers, built to verify this paper, which still seems to work.

EdgarVis

open · source

A 3D map of US public companies built from ~17 years of public SEC 10-K filings (~12k companies), built to visualize interesting clustering patterns and aberrations in businesses. Highly recommend rolling the dice next to the search bar to find outlier/interesting companies and the insights subpage for a more traditional set of visualizations.

Chain audio

open

Live USDC transfers across nine blockchains, turned into sound. Inspired by listen to wikipedia, a new and more natural way to understand this data.

MCP ingest