Hi there!
You’re invited to a meetup focused on specialists working with data engineering, analytics, and machine learning.
Join us at our office in Cyprus.
We’ll send detailed instructions on how to find us by email after you register.
Contact the organizers via Telegram: @space307team
Agenda:
17:30 — Meeting guests
18:00 — Opening remarks
18:05 – 18:45 — Bootstrapping Multilingual Datasets with LLMs
Speaker: Hassan Hammoud, Data Scientist at InDrive
18:45 – 19:25 — Autonomous AI Trading Agents: Evolution to ElizaOS
Speaker: Roman Abramov, AI Engineer at Space307
19:25 – 20:05 — Streamlining Helpdesk with RAG & LLMs
Speaker: Maksim Polyakov, AI&ML Developer at DataArt
20:05 – 22:00 — Networking
Talks & Speakers:
-
Bootstrapping Multilingual Datasets with LLMs
Hassan Hammoud, Data Scientist at InDrive
This talk explores how to effectively leverage Large Language Models (LLMs) in building multilingual datasets, with a focus on techniques that maximize LLM output quality—such as step-back prompting, chain-of-thought generation, and structured output. We’ll examine how LLMs can replace traditional translation APIs and assume roles typically filled by human annotators, streamlining the data creation process. The session presents a complete data generation pipeline, along with sampling strategies that improve linguistic coverage and data diversity, offering a scalable approach to training language models in low-resource or emerging languages.
-
Autonomous AI Trading Agents: Evolution to ElizaOS
Roman Abramov, AI Engineer at Space307
This talk explores the development of autonomous AI agents designed to analyze cryptocurrency markets through news sentiment and technical indicators. We’ll examine how these agents generate actionable trading signals for users, enabling them to select personalized trading setups aligned with their strategies. The presentation will focus on our migration to ElizaOS—a framework that enhances agent autonomy and expands distribution capabilities beyond our proprietary application to various social media platforms. We’ll discuss the architectural considerations, implementation challenges, and unique features of ElizaOS that enable truly autonomous agent operation. The session presents our migration journey, highlighting how this evolution creates more responsive, independent trading agents capable of broader market reach and enhanced user engagement across multiple channels.
-
Streamlining Helpdesk with RAG & LLMs
Maxim Polyakov, AI&ML Developer в Data Art
We’ll dive into the practical applications of RAG and LLMs, using the Global Helpdesk project as a compelling case study. We ran into the challenge of unifying countless email-based helpdesks into a single system while drastically reducing incoming requests by leveraging internal Confluence resources. We’ll explore how tackled this challenge, mitigated LLM hallucinations, and optimized the system to improve search quality and overall efficiency.
By building a classic RAG system on Confluence’s public spaces, we encountered issues like outdated or conflicting data, the limitations of basic text chunking for vector databases, and the need to balance user-specific context with global company queries.
