GenAI Cracow #13 - Cost Optimization
Welcome to GenAI Cracow (Lucky) #13
Agenda
LLM Cost Optimization at scale by Adam
Managing cost in Generative AI Projects - Discussion panel
Networking & EPAM XR+GenAI demo
Abstracts
LLM Cost Optimization at scale
Adam will discuss strategies for reducing LLM costs as part of the production-readiness journey. He will cover key approaches like prompt caching and batch processing, focusing on OpenAI and Anthropic models. He will share the results of cost-cutting experiments on a high-traffic GenAI solution, showing how they achieved major savings. You will gain insights to make your LLM deployments more cost-efficient without compromising performance.
Managing cost in Generative AI Projects
This panel with Konrad Bujak, Adam Krzysiek and Grigorij Dudnik will examine the real costs of GenAI implementations. Beyond LLM API expenses, we'll discuss infrastructure, databases, staff requirements, data acquisition, engineering work, and other hidden costs. Panelists will share practical strategies for managing total costs in GenAI projects based on their implementation experience.
EPAM XR+GenAI demo
One of EPAM’s demo projects, LuxAR Estates. Using Apple Vision Pro, you can virtually tour homes, interact with an AI-driven consultant for instant property insights, and experience properties in immersive AR. This AR+AI solution shows how technology can transform home searching, making it more interactive, efficient, and engaging.
Speakers
Adam Krzysiek
Engineering Manager at EPAM with experience in Python Web Development and Python ML across industries including Energy, Retail, Media, and Telecom sectors. For the last two years focused on GenAI, supporting GenAI business development and leading both rapid prototyping and development of production-grade solutions. For the last four years, teaching Python and GenAI to master's degree students as academic trainer at Politechnika Krakowska and AGH in Cracow.
LinkedIn
Grigorij Dudnik
Grigorij created Clean Coder, an agentic framework for autonomous coding by AI that he used to build his startup takzyli.pl where he's CTO and co-founder. He also developed TinderGPT, an open-source AI dating assistant, and built a Raspberry Pi robot that makes complex decisions using open-source LLMs. His expertise includes fine-tuning models, and his talks have been recognized as "Hot" at DevAI Warsaw and won an award for most content-rich speech at AI Expo Europe.
LinkedIn | GitHub
Sponsors
EPAM Systems is a leading provider of digital transformation services in the world and a leader in the development of digital and physical products and digital platform engineering services. Through its innovative strategy; integrated advisory, consulting, and design capabilities; and unique 'Engineering DNA,' EPAM's globally deployed hybrid teams help make the future real for clients and communities around the world by powering better enterprise, education and health platforms that connect people, optimize experiences, and improve people's lives.