Lorenc Zhuka

Co-founded a startup that builds semantic search engines and RAG-based private chatbots enabling users to quickly find documents and obtain clear answers from knowledge bases. Works with large language models, private and external APIs, embeddings, and OCR to extract and structure information from PDFs for retrieval. Develops AI automations integrated into daily workflows to save time and improve efficiency for companies.

Developed state-of-the-art AI solutions focusing on NLP tasks and building RAG-based large language model applications. Built customized RAG pipelines with tailored data loaders, chunking methods and retrievers; created automatic data generation and evaluation pipelines for RAG; hosted open-source LLMs with various frameworks like vllm and text-generation-inference or self-host; developed chatbot applications with React and FastAPI.

Contributed to the development of an AI agent that searches across several databases to locate relevant research papers and generate evidence-backed responses with paper references using GPT-3.5-turbo. Built autonomous agents based on the BabyAGI architecture and the ReAct framework, enabling self-learning, reasoning, and decision-making in process development contexts. Applied machine learning within process development; built AI automation workflows.