Research & models / 2026
CSI LLM Lab.
A private research track for customer-service scoring and structured extraction models.
Research & modelsCSI LLM LabSystems shaped for people.
01 / Context
The problem behind the work.
Smaller models can be attractive for cost and control, but their quality must be compared on the actual service task.
02 / Contribution
What I built.
- 01
Fine-tuning and evaluation experiments using quantized model loading, LoRA/PEFT and structured datasets.
- 02
Cloud Run GPU capacity and private-checkpoint work to understand operational feasibility.
03 / Where it stands
The research is included as a technical case study. No public product screenshot or model endpoint is claimed.
Technology and methods
GemmaUnslothLoRA / PEFTTRLTransformersPyTorchHugging FaceCloud Run L4