оffline

From Hype to Reality: Building & Scaling Practical LLM Solutions

When and where:

February 29th, 17:00

Registration deadline:

February 28th

Format:

offline in Kyiv

Price:

donation 500+ UAH

pc

Objective:

Large Language Models have gained popularity across industries, but moving into practical application requires a shift both in mindset and skillset. This meetup delves into building and scaling real-world LLM solutions.

Equip yourself with the knowledge and strategies to transform LLM hype into tangible results. Join us and be part of the conversation shaping the future of LLM adoption.

Who is it for:

  • AI/ML Engineers and Specialists with 3+ years of experience in building and deploying AI/ML models
  • Developers working on integrating AI solutions into product pipelines and infrastructure
  • Data Scientists/Analyst interested in understanding how data interacts with LLMs

To participate:

  • Register
  • Check your email and find the link for the donation
  • Take a laptop or notebook with you, to write down important insights

What to expect:

  • Knowledge exchange and case study review
  • Networking and connection with fellow AI/ML experts passionate about driving LLM innovation
  • Discussion of key industry trends and challenges
  • Broader understanding of LLMs commercial potential
  • Understanding of advanced techniques for tackling key challenges like bias, explainability, and trust in LLM systems
  • An opportunity to ask questions and get answers

Speakers:

Mariana Romanyshyn
Mariana Romanyshyn

Computational linguist, Area Tech Lead at Grammarly

  • Mariana has professional experience in developing natural language processing (NLP) applications, ranging from low-level solutions like syntactic parsing to features like sentiment analysis and fact extraction.
  • For the past nine years, Mariana has been leading the team of computational linguists at Grammarly, where they develop error correction and text improvement algorithms. She is passionate about computational linguistics and NLP in Ukraine, which is why she co-organizes the UNLP workshop.
  • An educator by calling, Mariana also enjoy organizing summer schools, teaching NLP courses, and supervising diploma papers. She has enjoyed exploring the boundaries of large language models for the past year, but my primary interest still lies in structural linguistics as a method of formalizing the natural language.
Roman Kyslyi
Roman Kyslyi

Applied Data Scientist, AI HOUSE Podcast Host

  • Roman is a Data Scientist with over 6 years of experience in various fields, primarily in NLP.
  • He also teaches at KPI.
  • He is the host of the AI ​​HOUSE podcast.
  • Сan assist with: NLP, LLM, parsing and data collection.
Illia Strelnykov
Illia Strelnykov

Data Scientist at YouScan

  • 4 years at YouScan, mainly working on NLP.
  • ~2 years at QuantuMobile. NLP/CV/General ML.
  • 1 year at GalacticExchange (now Skale) – Swiss Army Knife.

Agenda

5:00 — 5:30pm
Registration & networking

5:30 — 6:30pm
Topic: Practical Advice on Building Sustainable Prompt-Based Solutions
Designing a GenAI feature based on a single prompt is an easy-peasy task. It's when you have a hundred prompts their maintenance becomes challenging. In this tech talk, we'll delve into the low-level elements of building prompt-based solutions. We'll discuss prompt templating, prompt evaluations, prompt iterations, and other practices that are frequently overlooked yet can make the logic of your system sustainable and maintainable. This talk will equip you with the knowledge and skills to transform an overwhelming task into an organized, manageable process.

Mariana Romanyshyn

6:30 — 7:10pm
Topic: Creation and development of Insights Copilot
Insights Copilot is the first social media listening assistant powered by ChatGPT to find qualitative insights from millions of online conversations. Ilya will present his teams' experience building Insights Copilot, delving into the limitations they encountered during development and the strategies to overcome them. The talk will cover technical aspects and provide insights into the product development process, offering a broad understanding of the project.

Ilya Strelnikov

7:10 — 7:20pm
Break

7:20 — 8:00pm
Topic: Building and Evaluating LLM Agents
Everyone knows how to do RAG applications. But usage of LLM Agents can significantly increase value from LLM applications. This workshop, will show you how to develop, evaluate, and iterate LLM Agents, using ReAct (Reasoning and Acting). In this workshop, you will learn: what is an Agent; what is ReAct; how to use to build LLM Agent (with LamaIndex); how to evaluate your LLM Agent using open source LLM Observability tools like MLFlow; how to iterate your way to an effective app that will be approved for production and what to do when agent fails.

Roman Kyslyi

8:00 — 9:00pm
Cocktails & networking

FAQ:

Menu
Join us
Ua