Важлива новина! В цю п'ятницю 20 грудня в 18:30 в тімсі відбудеться 79 зустріч нашої групи! Буде мега доповідь та, як завжди спілкування після неї!
Чекаємо на вас! Детальний опис доповіді нижче.
Посилання на ІВЕНТ.
Topic:
Типові проблеми у розробці та тестуванні сучасних AI застосунків. Майже свята, тож самий час поспілкуватися формально та не формально на афтепаті :).
This session is about the new problems that developers can encounter with new LLM-based projects. Some problems are easy to solve with a proper design, evaluation strategy, and a few experiments, like adjusting search indexes and tuning for optimal data retrieval. Others are more serious, like the difficulty of keeping context in large token windows, automating prompt engineering, formatting output and finding the right balance between performance, cost, and complexity.
Key topics
- Making the right choices with design and components of modern AI apps.
- Response evaluation problem with auto and human evaluation.
- Multi-agent/LLM approach for validation and secondary fact-checking.
- Finding a balance with RAG data ingestion of different document types, repeatable information, and optimal search.
- Challenges with Model and RAG poisoning by external actors and company employees.
Join this talk to learn more about these problems, design choices, strategies to minimize their impact, and how to prepare for your next LLM project.
Stanislav Lebedenko is a cloud architect working at Solidify AB. Co-organizer of Odesa Microsoft. NET User Group, speaker and author, an active member of international and Ukraine .NET/Azure communities. He has seventeen years of experience as a developer, Microsoft Azure Most Valuable Professional since 2020, Microsoft Certified Trainer, and Certified Azure Solution Architect Expert.
