Java and AI Mini Conference - Java User Group Macedonia

20 December 2025, Netaville
starting at 09:00

Agenda

Time	Talk	Presenter
9:00 - 10:00	Registration and coffee
10:00 - 10:40	Unified In-Context Learning: Integrating Raw Telemetry and Natural Language Directives for Zero-Shot Forecasting	Gjorgji Madzarov & Dimitar Peshevski
	Traditional time series forecasting requires complex statistical pipelines, specialized coding, and isolated data environments. This presentation details the initial findings of an exploratory research effort into Instructionable Time Series Analysis, a novel approach leveraging the in-context learning capabilities of Large Language Models (LLMs). We present early evidence demonstrating how raw temporal data and natural language directives can be unified within a single interpretation. This initial success suggests a strong potential for allowing non-experts to execute sophisticated tasks, such as data imputation and multi-step forecasting through conversational directives, which could significantly reduce the time-to-insight in data analysis workflows.
10:40 - 11:00	Break
11:00 - 11:40	AI-driven reverse engineering	Martin Toshev
	In this session we will reveal how can AI help us analyse better code repositories allowing us to speed up the process of reverse engineering an unknown codebase.
11:40 - 12:00	Break
12:00 - 12:20	Gen AI with local models and LangChain4j on 2 x NVIDIA Sparks	Kristina Stolikj & Dimitar Peshevski
	This talk demonstrates a fully local GenAI development setup, presented live on two physical NVIDIA Spark machines, running offline LLMs such as GPT-OSS and Qwen models. We show how the two Spark PCs are networked on-site to form a small local inference environment, covering practical configuration choices, hardware constraints, and common pitfalls of GPU-based local development. In the second part, we showcase how LangChain4j can be used with local models to implement prompting, Retrieval-Augmented Generation (RAG), and simple agent-style workflows. All demos are executed live on the two NVIDIA Spark machines, giving the audience a concrete, hands-on view of what local GenAI development looks like in practice.
12:20 - 12:30	Short break
12:30 - 12:50	Implementing Human-In-The-Loop with LangGraph4j	Ivan Kichukov
	Inspired by Python’s library LangGraph, the Java library LangGraph4j offers the opportunity to build stateful, multi-agent applications with Large Language Models (LLMs). This talk will highlight how LangGraph4j’s graph-based workflow engine makes it possible to combine LLM steps with human review.
12:50 - 13:00	Short break
13:00 - 13:40	Java and GenAI: from basics to enterprise	Panche Chavkovski
	Java GenAI libraries are getting more mature as every day passes by. A lot of examples and documentation are available, but the business needs are on a different level. This talk will show the path from starting integration towards enterprise ready implementations.
13:40 - 15:00	Beer and networking

Speakers

Gjorgji Madzarov

Bio:

Gjorgji Madjarov is a Professor at the Faculty of Computer Science and Engineering, “St. Cyril and Methodius” University in Skopje. His active research interests are on designing advanced approaches for automated time-series modeling and forecasting, stream processing and AI Operations using an extensive set of state of the art algorithms. Building on more than 15 years of research, development and management experience, he now leads a highly skilled team of professionals, dedicated to delivering state-of-the-art ML solutions to the world.

Dimitar Peshevski

Bio:

Dimitar Peshevski is a dedicated researcher and a PhD student in Data Science in Computer Science and Engineering at the Faculty of Computer Science and Engineering. He currently serves as a Demonstrator, having previously held the role of Lab Assistant for courses including Introduction to Data Science and Operating Systems. His scientific and research interests are focused on Data Science, Machine Learning, Artificial Intelligence, Parallel and Distributed Processing, and Software Engineering. Dimitar has authored and co-authored multiple scientific papers presented at both domestic and international conferences and published in academic journals.

Martin Toshev

Bio:

Martin is a solution architect and IT consultant conducting professional trainings for novice and experienced developers. He is a Java enthusiast and one of the leads of the Bulgarian Java User group (BG JUG) where he is one of the organizers of jPrime conference. He is also one of the organizers of ChurConf Java community conference in New Zealand.

Kristina Stolikj

Bio:

Software Engineer at Sorsix with five years of experience building fullstack applications, from Java roots to Kotlin expertise. Passionate about technology, currently developing medical software that streamlines workflows for doctors and improves patient outcomes. Always exploring new technologies and better approaches to solve real-world problems.

Ivan Kichukov

Bio:

Ivan Kichukov is a software engineer at G+D Netcetera. The greater part of his professional focus has been the back-end side of things when building web-applications in the Spring Boot + Java ecosystem.

Panche Chavkovski

Bio:

Pance is a Principal Software Engineer at G+D Netcetera and the current JUGMK leader. He is a Java developer on daily basis, GenAI enthusiast and part-time hardware maker. Pance is known as the creator of the pulse.eco platform.