AI-Powered Technical Solution Recommender (Generative AI Project)

Developed a generative AI recommender system that provides relevant design solutions based on user-described technical problems. Applied transformer models (BERT, LLMs via Hugging Face) for semantic analysis, chunking, and embedding generation from unstructured text.

Embeddings were indexed using internal vector database and matched via a vector database to enable fast and accurate semantic search.

The system returns one or multiple solutions with correlation scores, supporting informed decision-making. The model was benchmarked and outperformed previous systems in recommendation accuracy and semantic relevance.