The next frontier in GenAI: Building LLM Apps

At our recent FridayTalk, our CTO, Jovica Zorić, delivered an engaging presentation on Generative AI (GenAI) and Large Language Models (LLMs), sharing his hands-on experiences with these technologies.

He began by explaining the basics, focusing on retrieval-augmented generation (RAG), a technique that allows models to generate contextually relevant text and access necessary information effectively.

Jovica discussed the importance of vector databases for managing large datasets and how embeddings transform natural language into numerical vectors for improved model understanding. He highlighted tools like Chainlit and Streamlit, which facilitate the creation of user-friendly and interactive interfaces. Additionally, he demonstrated how writing an API wrapper layer on top of LLM generation can integrate these solutions into other UI tools and services.

He introduced several useful frameworks and tools, including FAISS, ChromaDB, Weaviate, Haystack, HuggingFace, LangChain, LlamaIndex, as well as monitoring tools like LangFuse. Jovica emphasized that understanding these tools is important for building effective RAG applications. While these are the tools he has tried, many others are available; the field is growing. Looking ahead, he mentioned topics like prompt engineering, deployment strategies, and multilingual embeddings. He also highlighted cloud platforms and their AI offerings like AWS Bedrock, Azure AI, and Watson AI, which he plans to explore further.

Despite the challenges, Jovica’s journey demonstrates just how fast and exciting the AI field is, offering numerous opportunities for innovation and discovery.

Explore our blog section for upcoming captivating topics that we’ll be sharing with you soon.