LLM Engineering

Developed a solution to predict product prices from short descriptions using Frontier models.

• Conducted model comparisons on a dataset of 200 training samples and 50 validation samples.

• Models tested include: Random & generator, average price, feature engineering, bag of words, Word2Vec, support vector regression, random forest, human guess, GPT-4O mini, GPT-4O, and Claude.

• Fine-tuned the testing dataset with the frontier model to achieve improved result.