Voxxed Days Bucharest 2024 – Hands-on Lab: Coding Context-Aware RAG Applications for Enterprise-Level (ContinualBot, 2024)
Info
- Project type Public Speaking, Tech Presentation
- Date 29th of March 2024
- My Role Speaker, Mentor
- Topics AI, Java, Python, Chatbots, NLP, NLU, Context handling, Knowledge representation, Similarity, Semantic search
- Keywords Conversational AI, Natural Language Processing and Understanding, Telegram Bots, Spring AI, LLM, Qdrant
- Skills developed Presentation, Speaking
- Event details
Description
In the fast-changing world of technology, creating conversational agents that excel in large business environments is key to smarter and more effective digital help.
AI Chatbots cover a broad spectrum of use cases, such as impersonations, customer support, assistants, and even substituting parts of GUIs. Chatbots are likely to be instrumental in transforming our current approaches to UX and privacy.
In this workshop, we will build from scratch such a system by emphasizing the needs of large enterprises.
Live-coding session agenda:
- Introduction to Neural Networks, GenAI, Conversational Agents Architectures and RAG.
- Data preparation: Streamlining Unstructured Data.
- Natural Language Understanding: Setup LLM and NLP toolkits.
- Knowledge Representation: Data Embedding and Chunking.
- Retrieval: Semantic Search and Reasoning.
- Privacy: Strategies for implementing robust privacy policies.
- Edge Cases Discussion: Fine-tuning, Prompt Engineering and Knowledge Graphs.
- Q&A
This workshop aimed to equip developers with the knowledge and tools to build advanced, privacy-aware, and efficient AI chatbots.
In terms of preparation, Voxxed Days Bucharest was one of the most complex technical events I've ever tackled.
The challenge wasn't just delivering a speech with slides; it involved a two-hour live coding session. During this session, I programmed a functional MVP in real time and also provided support to the participants who were coding alongside me. In the background, while writing code, I explained the underlying theoretical concepts, which necessitated the use of slides.
Furthermore, I had to prepare to manage an entire room simultaneously, aiming not only to ensure that attendees finished with a working application but also to help them understand what they were doing.
This undertaking required extensive preparation: I created a step-by-step wiki for attendees to follow if they fell behind, and a Git repository to enable them to jump straight to the next steps if necessary.
I rehearsed everything the participants might do in those two hours, ensuring everything functioned on both Windows and Mac OS. A significant amount of effort was required behind the scenes, so I repurposed many components of this project from the Continual Bot.
I am grateful for this opportunity provided by Andra Ghibutiu and Alex Proca and am looking forward to further collaborations with Voxxed Days and the Incremental Community!
References
Tech stack
- Java, openjdk-19, maven 3.8.1
- Python 3.11
- Spring and the new Spring AI
- Telegram Bots API doc
- Telegram Bot, Java lib
- Docker
- Qdrant