This repository contains the training pipeline, datasets, evaluation harness, and model artifacts for a domain-specific LLM focused on financial risk, model governance, and compliance-aware reasoning.

Scope

Instruction-tuned LLM for financial risk & governance
Conservative, regulator-friendly tone
Structured outputs (controls, RACI, checklists, evidence)

Status

Local smoke training (Apple Silicon / MPS): ✅
Cloud QLoRA training (7B+): planned

Repository Structure

data/ – raw and processed datasets
scripts/ – training, inference, evaluation
src/ – prompts and formatting logic
eval/ – evaluation prompts
docs/ – model card and documentation

swnema/swati-finriskgov-llm-smoke

swati-finrisk-llm-v1

Scope

Status

Repository Structure

swnema/
swati-finriskgov-llm-smoke