DataFest — Belgrade, May 31

FON, University of Belgrade

---

12:00 — Guest Registration & Welcome

13:05 — Opening remarks · Salavat Gariffullin, ODS Serbia

13:15 — Welcome Address

Prof. Dragan Vukmirović, PhD (FON University) From Big Data to AI‑native: 7V, Synthetic Data and the New Role of Data Science in Industry

---

---

🏛 Stage 0 · Читаоница · 140 seats

(Library, Ground Floor)

🤖 Industry & Robots

Time	Speaker	Company	Talk	Description
13:30	Fedor Kurdov	Yandex	RL for Real-World Robot Motion Planning	How RL without imitation learning was built from scratch for Yandex sidewalk delivery rovers.
14:00	Aleksey Postnikov	Sber Robotics Lab	Physical AI: Status and the Road Ahead	Broad overview of Physical AI: synthetic data, sim-to-real, RL over behavior cloning, and learning from human videos.
14:30	Dmitrii Iunovidov	LogicYield	Making Industrial CV Fly on Edge CPUs	Running industrial computer vision on edge CPUs in harsh factory conditions using inference optimization and neuro-symbolic methods.
15:00	Fedor Konovalenko	MIL Team (MIPT)	From Model Compression to Local Inference Platform	How a model compression tool evolved into a local GenAI inference platform with OpenAI-compatible API and multi-engine support.

---

15:30 — 16:30 · Break

---

🔬 Career & Quality

Time	Speaker	Company	Talk	Description
16:30	Maksim Artemev	Morphic	How (Not) to Find a Job	Common candidate mistakes, hiring mechanics, and salary negotiation from the perspective of applicant, manager, and recruiter.
17:00	Vladimir Kukushkin	Independent Researcher	Beyond Funnels: Advanced UX Analytics	How to study user behavior deeper than traditional funnels using advanced UX analytics tools and journey analysis.
17:30	Alexey Vasilev	Sber AI Lab	SplitLight: RecSys Evaluation Toolkit	Open-source toolkit for analyzing datasets and split strategies in RecSys to make offline evaluation transparent and reproducible.
18:00	Alexey Korotkov & Timofey Garaev	MIPT AI Institute	SHARP: Span-level Hallucination Annotation for Reasoning Paths	New span-level dataset for hallucination detection in LLM reasoning paths, yielding better downstream quality for PRM models.

---

---

🏛 Stage 1 · Амфитеатар 1 · 180 seats

(Lecture Hall, 1st Floor)

🛡️ Trust in AI

Time	Speaker	Company	Talk	Description
13:30	Oleg Sekachev	Yandex	Vector DB & Specialist LLM in Labelling Pipelines	Using a vector DB to reuse accumulated annotation data, plus a domain-tuned LLM pipeline as a cost-quality compromise.
14:00	Anastasiia Margolina	Banco Plata	How we (didn't) build an AutoEval	A story about evaluating AI when the answers involve real money — and how a single prompt turned into a full methodology.
14:30	Stefan Hačko	Foursquare	LLM-Powered Harmonization of 100M+ Places	How Foursquare uses LLMs and vector embeddings to clean, match, and unify massive third-party venue datasets at scale.
15:00	Michael Diskin	Independent Researcher	When Models Should Stay Silent	Measuring model uncertainty, calibrating confidence, and implementing rejection mechanisms for reliable LLMs in production.

---

15:30 — 16:30 · Break

---

⚡ Agents & LLMs

Time	Speaker	Company	Talk	Description
16:30	Ksenija Blažević	Lemana Pro	4 Anti-Patterns in Agentic AI	Common anti-patterns in agentic AI and how to replace them with leaner, cost-efficient architectures.
17:00	Dmitrii Krasnov	Zencoder	Orchestrating Coding Models	Comparison of sequential, parallel, and OSS orchestration for coding models and their impact on SWE-bench-like benchmarks.
17:30	Nikita Severin	Independent Researcher	Knowledge Transfer from Pre-trained LLMs to Recommender Models	Efficient knowledge transfer from LLMs to recommender models without costly serving-time inference or architectural changes.
18:00	Ivan Bushmarinov	Perplexity	User-Guided LLM Answer Quality Evaluation	Leveraging thread-style user feedback and small trained models to evaluate frontier LLM answers and enable scalable benchmarking.

---

---

🏛 Stage 2 · Амфитеатар 2 · 180 seats

(Lecture Hall, 2nd Floor)

💳 Ranking & Banking

Time	Speaker	Company	Talk	Description
13:30	Alexander Eroshenko	Yandex	LLM-Powered Item-to-Item Recs in Lavka	Practical case of deploying a compact LLM (Gemma ~270M) for item-to-item recommendations of substitutes and complements.
14:00	Boris Tseitlin	Banco Plata	Learning from Unstructured Sequences in 2026	Overview of self-supervised and foundation-model approaches to embeddings from transactions, events, and unstructured sequences.
14:30	Mikhail Sysoev	Banco Plata	PV Models in Retail Lending	PV models and approaches to optimizing product parameters in card-based fintech products.
15:00	Victor Barbarich	Banco Plata	Transformers Replace Feature Engineering in Scoring	Moving from manual feature engineering in credit scoring to transformers that learn directly from raw account and employment histories.

---

15:30 — 16:30 · Break

---

🎙️ The Voice of AI

Time	Speaker	Company	Talk	Description
16:30	Aleksandr Nikolić	Independent Researcher	Borealis — how to train an audio LLM for the price of a MacBook	Practical story on training an audio LLM with a limited budget using the example of Borealis.
17:00	Pavel Mazaev	Yandex	Device-Directed Speech Detection for Alice	Production system for detecting speech directed at Yandex Alice to enable natural dialogue without a constant wake word.
17:30	Ilya Shigabeev	Langswap.app	TTS in 2026: Open Source vs Big Tech	Practical view on modern TTS modeling, competing with big tech, and why small open-source models still matter.
18:00	Pavel Guliaev	Independent Researcher	Video2Text: Industry State & Practical Choices	State of the Video2Text industry — what works, current limitations, and how to pick solutions for production load and budget.

---

---

18:30 — Closing remarks · Salavat Gariffullin, ODS Serbia

18:45 — After-party 🎉