Research background
The NLP research behind my applied LLM work: how meaning is represented, compared and retrieved.
My graduate research at IASBS focused on scoring semantic similarity between augmented sentence pairs, using autoencoders and a Siamese BERT architecture benchmarked against SBERT and Augmented SBERT baselines. It shaped how I think about embeddings and evaluation.
Today I apply that foundation to production-style LLM systems: RAG pipelines, MCP-based agents and Neo4j-backed semantic search.
Semantic Similarity & Scoring
Autoencoder and Siamese BERT models for scoring augmented sentence pairs on semantic textual similarity.
LLM Applications & Agents
Prompt engineering and optimization, RAG pipelines, and AI agent design (via LangChain, MCP).
Graph-Based Retrieval
Neo4j graph databases for semantic search in LLM applications.