mirabdullahyaser/LLaMA3-Financial-Analyst

Implements a two-stage architecture: LoRA-adapted Llama-3-8B trained on curated 10-K Q&A data using the Unsloth framework with 4-bit quantization, paired with FAISS in-memory vector storage and BAAI/bge-large-en-v1.5 embeddings for semantic retrieval. Integrates the SEC API for real-time 10-K document ingestion, processing Risk Factors (Section 1A) and MD&A (Section 7) specifically, while using LangChain to orchestrate the RAG pipeline that contextualizes queries before inference.

No commits in the last 6 months.