Andrei Beliankou
I am the Technical Lead for Data & AI in the Energy Retail Team at E.ON Digital Technology. Unofficially, I describe myself as a Software Engineer with a Data affinity. I write both code and texts.
Being a member of the E.ON GenAI Core team, I've been developing Generative AI solutions for different business units within the E.ON family. Multilingual Search, LLMs, Agentic RAG, Knowledge Graphs - a small selection of buzzwords for our daily activities.
Machines listen to me in SQL, Python, Ruby, AWK, Bash, YAML and (hopefully very soon) Rust. My beloved machines mostly live in the Azure Cloud.
I am passionate about building technical teams around a goal and having fun crafting solid software.
You can talk to me in German, English, Russian, Polish, Ukrainian, Belarussian, Italian, and Spanish. Latin may also be worth trying.
I still strongly believe in non-dumb statistical approaches to AI. The next step will only be possible through a combination of the humanities and science.
Session
Keywords: Explainable AI, enhanced RAG, GraphRAG, LLMOps, dialog system evaluation.
Designing reliable dialog flows for LLM-based systems remains challenging once conversations require branching, correction, or multi-step reasoning. Dialog graphs often evolve organically and accumulate structural issues: endless correction loops, dead subpaths, redundant validation steps, overly generic catch-all branches, or linear sequences that should be collapsed. Such phenomena raise operational costs, significantly increase TTFT and make the system answer less predictable and explainable.
Many solutions try to introduce an all-fit generalized RAG retrieval solution. Contrary to this, we present our empirical learnings on how to enhance system speed, lower overall costs and offer a better dialog graph explainability through enhanced LLM call tracing and iterative enhancements for common dialog paths.
We also show that more elaborated knowledge retrieval strategies like GraphRAG may drastically enhance overall response quality and shorten the dialog graph. We evaluate several approaches and give recommendations on how to leverage more complex document indexing phases for inference time benefits.
Overall, the session argues that scalable conversational systems require not only better prompts, but explicit graph structures paired with rigorous tracing and data-driven optimization.