— Data Sanity Talks Belgrade Oct 4: A talk recording —

🎙 When LLMs Train Your Models: Using AI to Build Better AI

Speaker: Stanislav Tsepa, ML Engineer @ FAANG

We'll learn how AI can help create smarter AI — from cleaning noisy datasets and prompt-optimization to generating synthetic data and distilling knowledge into smaller models.

📺 Watch here: https://youtu.be/uKTuARCM6EA

— Data Sanity Talks Belgrade Oct 4: A talk recording —

🎙 Building a Data Platform as a Single Source of Truth

Speaker: Dušan Zamurović, Engineering Manager @ Foursquare

We'll explore and analyze a real-world journey of building an AWS-based, event-driven data processing platform that matured from backend infrastructure into a company-wide self-service hub.

📺 Watch here: https://youtu.be/9zJfetEkpfs

P.S. Only one talk left!

— Data Sanity Talks Belgrade Oct 4: A talk recording —

🎙 A Practical Guide to LLM Context Engineering

Speaker: Aleksandar Anđelković, Principal Engineering Manager @ Microsoft

Let's design a practical framework to overcome LLM context limits by selecting, compressing, and structuring information with RAG, summarization, memory, and agentic workflows to build smarter, more reliable AI applications.

📺 Watch here: https://youtu.be/k1HplsKzZ0M

————————

This is it! Huge thanks to all our speakers for the time and effort in preparing their talks <3

See you at our next event!

— Data Sanity Talks Belgrade Oct 4: A talk recording —

Here is one talk that we forgot to post saved for last!

🎙 Launching AI in Conservative Corporations: Turning Skepticism into Value

Speaker: Mariya Zlenko, AI Product Director @ Uralsib

Mariya shares her experience of bringing AI to a place where people are sceptical about the new technology, and what she learned along the way.

📺 Watch here: https://youtu.be/6Srf6iAfkX8

————————

P.S. We are about to announce two events in December — a seminar in Novi Sad and a workshop in Belgrade. Stay tuned!

🚀 One Seminar — Two Topics to dive in with Vadim Opolski (Senior Data Engineer @ DXC)!

📍 Novi Sad, Startit Centar
🗓 December 13, Saturday | 11:00 - 13:00
✒️ Register here! (free)

​At this seminar, Vadim will guide us through two interconnected storylines, each uncovering a different side of modern engineering:

​⚡️ Topic 1 — Architecture Problems and Solutions in Streaming AI Pipelines
Vadim will break down a production-grade streaming architecture for monitoring text, photos, and video using Spark and Kafka — highlighting pitfalls, design decisions, and practical solutions.

​⚡️Topic 2 — From TDD to SDD with AI Agents We’ll explore how Test-Driven Development is evolving into Spec-Driven Development with AI tools. Together we’ll test agents live, generate specs, create architecture and tests automatically, build a working prototype from scratch.

​For those who want to try everything live during the session, please make sure you have:
1. ​Any console-based AI agent installed (e.g., Claude Code, GitHub Copilot).
2. ​The SDD tool spec-kit: https://github.com/github/spec-kit.

​One seminar, two topics — both hands-on and packed with insights!

ℹ️ Novi Sad | Dec 13 | Free registration

The Data Sanity seminar in Novi Sad is already tomorrow!

🍪 Cookies are limited — register here for free to reserve yours!

Data Sanity Workshop: Optimizing AI Agent Prompts with DSPy

🗓 December 20, 11:00 - 13:00
📍 Belgrade, Nušićeva 15
🎟 https://luma.com/xi58ggeo

​In this hands-on workshop, you’ll be introduced to DSPy, a practical framework for optimizing prompts for AI agents.

Together with Michael Pavlukhin, we will:
🚀 Revisit the fundamentals of AI agents
🚀 Learn how to improve prompts using the GEPA evolutionary optimizer with DSPy
🚀 Build a movie recommendation agent
🚀 Experiment and gain hands-on experience

​Requirements:
🤖 A laptop
🤖 Proficiency in Python and notebooks
🤖 (Nice to have) Experience working with LLMs via APIs

✒️ Register for free to secure your spot!

​This event is hosted by Mainstream — the leading cloud provider in Southeast Europe.

The workshop begins in 24 hours!

A few tips:
❗️ Please register using your full name (required for lobby security)
💻 No prior experience with any specific LLM API is required
💻 We’ll be working in Google Collab, so there’s nothing to install ahead of time
💻 If you don’t have a laptop, you’re still welcome to attend and experiment with the notebook later at home

✒️ Register for free to secure your spot!

🚨 OpenTalks.AI: Last call for speakers!

Today is the final day to submit a talk proposal for OpenTalks.AI, an ML conference happening in Belgrade on Feb 19-20.

Yes, it’s a very late notice — but there’s still time!

Apply here: https://opentalks.ai/en/call-for-proposals

P.S. We, the Data Sanity team, are also planning our own event on the weekend following the conference — stay tuned!

Fellow data scientists and enthusiasts,

We, the Data Sanity team, are preparing a course on LLM and AI agent basics — covering both theory and practice (Python + LangChain & LangGraph ecosystem). We'll teach you fundamentals and more advanced topics, highlight main trends in the industry, and help you build an AI agent to practice your newly acquired skills.

We'll host 8-9 in-person seminars in Belgrade in March and April. But when exactly? You can help us decide by voting in the questionnaire below!

P.S. We are currently looking for a sponsor to make this course 100% free. If you'd like to sponsor and help the local data community — please reach out!

Hello data fellows,

OpenTalks.AI conference program is live!

The OpenTalks.AI 2026 conference will take place on Feb 19-20 at Sava Centar. There will be 3 tracks — research, engineering, and business — with hot topics like LLMs, CV, AI agents, compute optimization, and more!

🎟 Use promo code DataSanityOT26 to get a 20% discount on registration.