AIINFRA 201: Model Adaptation — Fine-Tuning & Quantization — Outcomes & Rubrics

Course Learning Outcomes

By the end of this course, students will be able to:

1. CLO1 — Evaluate a task and choose correctly among prompt engineering, retrieval-augmented generation, and fine-tuning — justifying by cost, data, and hardware.
2. CLO2 — Build, clean, and format instruction/chat datasets and run supervised and parameter-efficient fine-tuning (LoRA, QLoRA, PEFT) on a single GPU.
3. CLO3 — Apply preference tuning with DPO and track experiments while avoiding overfitting and catastrophic forgetting.
4. CLO4 — Quantize models into GGUF, AWQ, GPTQ, and FP8 formats and analyze the precision, quality, and memory tradeoffs.
5. CLO5 — Assess licensing and ethical constraints on model weights and data, and package an adapted, quantized model for serving.

Together, these outcomes ladder up to the certificate's program-level outcomes by building the practical AI infrastructure and architecture skills — model adaptation, optimization, and deployment-ready packaging — that California employers expect from job-ready AI infrastructure practitioners.

Rubrics

1. Lab / Hands-on Assignment Rubric