How to build your own Custom AI Assistant (Fine-Tune LLM Tutorial)

A walkthrough tutorial on fine-tuning a custom LLM using the FPT AI Factory platform. Using a Jupyter notebook environment with an H200 GPU, the process covers preparing Q&A training data, formatting it as JSONL, and fine-tuning the Mistral model with LoRA adapters to create a domain-specific interview coaching assistant. The tutorial demonstrates that LoRA allows efficient fine-tuning by adding a small expert layer rather than retraining the full model, keeping costs and time low. FPT AI Factory is the sponsored platform providing $100 in free credits for users to follow along.