LLM Engineering

Fine-tuned open-source models to outperform Frontier models in price prediction tasks.

• Used Qwen2.5 4-bit and Qwen2.5 8-bit models for testing.

• Refined and fine-tuned the model with a light weight testing dataset of 25K records.

• Achieved significant improvements, with Qwen2.5 outperforming Frontier models.