Tag: llm

All the articles with the tag "llm".

• Trained a Transaction Foundation Model from Scratch on 2×RTX 4090

  5 Jul, 2026

  Trained a 29M-parameter Llama decoder on financial transaction data from scratch on 2×RTX 4090s. Extracted embeddings and got +32% lift on fraud detection with XGBoost. End-to-end pipeline in 18 minutes.