Legal AI Assistant
Mar 2025 – Nov 2025
A RAG-based legal assistant for real-time, multi-turn conversations. Semantic retrieval in Weaviate with optimized embeddings, tokenization and document preprocessing, served by a FastAPI backend on AWS Bedrock LLMs, tuned through prompt engineering to 85–90% response relevance in evaluation tests.
- AWS Bedrock
- LLMs
- Weaviate
- FastAPI
- Python
- Embeddings
- Prompt Engineering
- LLM Evaluation
- RELEVANCE
- 85–90% in eval
- LLM
- AWS Bedrock
- RETRIEVAL
- Weaviate semantic
