Fine-tuning Mistral-7B using QLoRA to build a lightweight financial assistant that answers investment-related questions about Money Market Funds (MMFs) in Kenya. This project uses a small Alpaca-style dataset with 20 handcrafted examples to demonstrate local language model customization using PEFT techniques on Colab GPUs
# Mistral Kenya Finance QLoRA
Fine-tuning Mistral-7B on a small Kenyan dataset to build a financial assistant that answers questions about Money Market Funds (MMFs).
## đź’ˇ Objective
Use QLoRA (Quantized Low-Rank Adaptation) to fine-tune a 7B LLM on domain-specific financial data from Kenya. The final model answers common retail investment questions.
## 📊 Dataset
A small instruction-tuned JSON dataset (`kenya_mmf_data.json`) with 20 examples in Alpaca format:
```json
{
"instruction": "What is a money market fund?",
"input": "",
"output": "A money market fund in Kenya is..."
}