Check out the Multipass packages, with which you will gain simultaneous access to the speeches of as many as 3 conferences: DevAI by Data Science Summit, Python Summit and Ya!vaConf.
DevAI by Data Science Summit is an event for people involved in software development who want to stay up to date with trends and implement modern AI-based solutions. This is a unique conference dedicated to developers, supported by the largest Polish *data* community, which has been focused around the Data Science Summit for years.
An accompanying event is DevAI Expo - a job fair, products and services related to the conference theme.
Claude Code, Codex, Cursor, Copilot; vibe coding vs. structured engineering workflows; parallel task execution, repository management, code review automation and AI-native development pipelines.
Production reliability, maintenance, scaling, observability, evaluation systems, debugging, inference optimization, performance tuning and operational resilience for enterprise-grade AI systems.
Agent development patterns (ReAct, Plan-and-Execute, Reflection), memory architectures (short/long-term, episodic vs. semantic), tool use, MCP and function calling, distributed multi-agent coordination, shared state management, git worktrees and autonomous parallel coding inside the same repository.
Grounding strategies, context graphs, retrieval pipelines, context orchestration and retrieval optimization for agentic and LLM-based systems.
Testing nondeterministic systems, evaluation pipelines, LLM-as-a-judge methodologies, tracing, prompt injection mitigation, guardrails, sandboxing and reliability validation.
vLLM, quantization, on-prem deployments, edge inference, VLM pipelines, Document AI, TTS, STS and STT systems integration.
Token consumption management, inference cost optimization, model utilization efficiency, infrastructure economics and operational budgeting for AI workloads.
Organizational transformation in the AI era, developer productivity shifts, human-AI collaboration models, competency restructuring and AI-native engineering culture.
Governance, compliance, delivery accountability, AI adoption strategy, operational risk and long-term organizational adaptation.
At the event you will meet and listen to the best specialists from Poland and abroad.
Check Agenda 2026 - Selected by practitioners, the most interesting speeches on trends, latest solutions and production experiences.
08:30 - 09:00
Online Live
Official Opening
09:00 - 09:35
Online Live
Replacing the LLM in Your Agent with a Small Language Model: A Practical Guide
09:40 - 10:15
Online Live
Frozen at Training Time: Building Production Memory Layers for Stateful AI Agents
10:20 - 10:55
Online Live
Beyond Agents: A New Architecture for Reproducible Code Generation Workflows
11:00 - 11:35
Online Live
How to Keep Coding Agents from Breaking the Repository: Permission Boundaries, Verification Loops, and Human Control
11:40 - 12:15
Online Live
Deciding What Is "Correct": Verification as the New Role of the Engineer in the Agentic Era
12:20 - 12:55
Online Live
Agentic AI in Web3 Security: Autonomous Defenders and the Risks They Introduce
12:55 - 13:55
Online Live
13:55 - 14:30
Online Live
Agents That Never Stop: The Pattern Behind Self-Improving Systems
14:35 - 15:10
Online Live
Wirtualni pracownicy w SDLC: jak delegować pracę agentom AI bez utraty kontroli
15:15 - 15:50
Online Live
Vibes Don't Scale
15:55 - 16:30
Online Live
Co naprawdę psuje się w agentach AI na produkcji: pułapki decyzyjne, pozorna racjonalność i cichy dryf modeli
16:35 - 17:10
Online Live
Więcej niż Chat: Jak modele LLM rewolucjonizują inżynierię cech w tradycyjnych systemach Machine Learning
17:15 - 17:50
Online Live
MicroAgents: Applying Microservices Architecture to Production AI Agents
08:00 - 08:30
Centrum Nauki Kopernik
08:30 - 09:00
Room 1
09:00 - 09:35
Room 1
[KEYNOTE] When System Knowledge Disappears: How AI Helped Us Regain Control of a Project
09:45 - 10:20
Room 1
How AI Native development is reshaping the team structure
10:35 - 11:10
Room 1
Non deterministic tools applied to non deterministic process - code review harness @ Saily
11:20 - 11:55
Room 1
Agenci AI na straży faktów - generatywna AI w analizie danych i walce z dezinformacją
12:10 - 12:45
Room 1
Building software with AI in 2026: a practical playbook for protecting and monetizing what you ship
12:55 - 13:30
Room 1
Beyond the repository - agentic portfolio management
13:30 - 14:20
Centrum Nauki Kopernik
14:20 - 14:55
Room 1
Agentic coding w prawdziwym kodzie: gdy dostarczanie przestaje być wąskim gardłem
15:05 - 15:40
Room 1
We Have Met the Enemy and It Is Us: How AI Productivity Outpaces System Reliability
15:55 - 16:30
Room 1
What Coding Agents Still Get Wrong
16:40 - 17:15
Room 1
How a "Don't Touch It" Legacy App Became AI-First
17:25 - 18:00
Room 1
From Search to Booking: Building Trusted Android Workflows with App Functions
09:00 - 09:35
Room 2
Gdzie kończy się agent a zaczyna kod — niezawodność systemów multiagentowych na produkcji.
09:45 - 10:20
Room 2
Last Year's Frontier AI Now Fits on Your Laptop
10:35 - 11:10
Room 2
[KEYNOTE] Retrofitting Agentic RAG: The Missing NER Layer for Deterministic Recall. Finding and matching products across domains and millions of pages a week
11:20 - 11:55
Room 2
Vibe-Coding a Second Brain: What an AI Agent Actually Shipped vs the Demo
12:10 - 12:45
Room 2
How many agents can one GPU actually serve? Measuring the multi-user context wall before it hits production
12:55 - 13:30
Room 2
Talk With Your Data: Building AI-Powered Reporting
13:30 - 14:20
Centrum Nauki Kopernik
14:20 - 14:55
Room 2
The Extinct Attractor: What Happened to "I Don't Know" in Instruction-Tuned LLMs
15:05 - 15:40
Room 2
From Trace to Fix: Debugging Production AI Agents with Genie Code and MLflow
15:55 - 16:30
Room 2
Symbioza Predictive ML z LangGraph: wzorce architektoniczne, eksperci i pułapki ewaluacji
16:40 - 17:15
Room 2
Od alertów do diagnozy: architektura autonomicznego agenta SRE
17:25 - 18:00
Room 2
When Data can Act: Security Data Risks in LLM Systems.
13:30 - 14:20
Centrum Nauki Kopernik
14:20 - 14:55
Room 5
[KEYNOTE] Text-to-Diagram for AI Systems: Designing a DSL an LLM Gets Right
15:05 - 15:40
Room 5
Intent Before Intelligence: Fine-Tuning Small Models for Smarter Routing
15:55 - 16:30
Room 5
When AI Agents Fail in Production: Tracing, Budgets, and Tool Execution Ledgers
16:40 - 17:15
Room 5
Generative UI on Ollama Cloud: forcing a small open-weight model into legal moves
17:25 - 18:00
Room 5
AI end-to-end claim handling: od zgłoszenia szkody do błyskawicznej decyzji
10:35 - 11:10
Room RT
Autonomous Research w softdevie
11:20 - 11:55
Room RT
Co agenci AI muszą wiedzieć o naszych systemach - i skąd mają czerpać tę wiedzę?
12:10 - 12:45
Room RT
How to plan and control AI expenses
12:55 - 13:30
Room RT
How to use AI for code & design without generating slop?
14:20 - 14:55
Room RT
AI Agents - technicznie
15:05 - 15:40
Room RT
Kill Switch for AI: bare minimum or just the start?
09:40 - 09:55
Networking Zone
Ask a Speaker: [KEYNOTE] When System Knowledge Disappears: How AI Helped Us Regain Control of a Project
09:40 - 09:55
Networking Zone
Ask a Speaker: Gdzie kończy się agent a zaczyna kod - niezawodność systemów multiagentowych na produkcji.
10:25 - 10:40
Networking Zone
Ask a Speaker: Scaling Agentic Workflows: How We Built an MCP-Backed Engineering Operating Model
10:25 - 10:40
Networking Zone
Ask a Speaker: Last Year's Frontier AI Now Fits on Your Laptop
11:15 - 11:30
Networking Zone
Ask a Speaker: Autonomous Research w softdevie
11:15 - 11:30
Networking Zone
Ask a Speaker: Non deterministic tools applied to non deterministic process - code review harness @ Saily
11:15 - 11:30
Networking Zone
Ask a Speaker: [KEYNOTE] Retrofitting Agentic RAG: The Missing NER Layer for Deterministic Recall. Finding and matching products across domains and millions of pages a week
12:00 - 12:15
Networking Zone
Ask a Speaker: Vibe-Coding a Second Brain: What an AI Agent Actually Shipped vs the Demo
12:00 - 12:15
Networking Zone
Ask a Speaker: [TOPIC SOON]
12:00 - 12:15
Networking Zone
Ask a Speaker: [TOPIC SOON]
12:50 - 13:05
Networking Zone
Ask a Speaker: Building software with AI in 2026: a practical playbook for protecting and monetizing what you ship
12:50 - 13:05
Networking Zone
Ask a Speaker: How many agents can one GPU actually serve? Measuring the multi-user context wall before it hits production
12:50 - 13:05
Networking Zone
Ask a Speaker: [TOPIC SOON]
13:35 - 13:50
Networking Zone
Ask a Speaker: Talk With Your Data: Building AI-Powered Reporting
13:35 - 13:50
Networking Zone
Ask a Speaker: From Jobs to Workflows: Architecting the Human Layer of AI-Assisted Tech Teams
13:35 - 13:50
Networking Zone
Ask a Speaker: [TOPIC SOON]
15:00 - 15:15
Networking Zone
Ask a Speaker: AI Agents - technicznie
15:00 - 15:15
Networking Zone
Ask a Speaker: [KEYNOTE] Text-to-Diagram for AI Systems: Designing a DSL an LLM Gets Right
15:00 - 15:15
Networking Zone
Ask a Speaker: Agentic coding w prawdziwym kodzie: gdy dostarczanie przestaje być wąskim gardłem
15:00 - 15:15
Networking Zone
Ask a Speaker: The Extinct Attractor: What Happened to "I Don't Know" in Instruction-Tuned LLMs
15:45 - 16:00
Networking Zone
Ask a Speaker: [TOPIC SOON]
15:45 - 16:00
Networking Zone
Ask a Speaker: We Have Met the Enemy and It Is Us: How AI Productivity Outpaces System Reliability
15:45 - 16:00
Networking Zone
Ask a Speaker: Intent Before Intelligence: Fine-Tuning Small Models for Smarter Routing
15:45 - 16:00
Networking Zone
Ask a Speaker: From Trace to Fix: Debugging Production AI Agents with Genie Code and MLflow
16:35 - 16:50
Networking Zone
Ask a Speaker: Symbioza Predictive ML z LangGraph: wzorce architektoniczne, eksperci i pułapki ewaluacji
16:35 - 16:50
Networking Zone
Ask a Speaker: What Coding Agents Still Get Wrong
16:35 - 16:50
Networking Zone
Ask a Speaker: When AI Agents Fail in Production: Tracing, Budgets, and Tool Execution Ledgers
17:20 - 17:35
Networking Zone
Ask a Speaker: Generative UI on Ollama Cloud: forcing a small open-weight model into legal moves
17:20 - 17:35
Networking Zone
Ask a Speaker: How a "Don't Touch It" Legacy App Became AI-First
17:20 - 17:35
Networking Zone
Ask a Speaker: Od alertów do diagnozy: architektura autonomicznego agenta SRE
18:05 - 18:20
Networking Zone
Ask a Speaker: AI end-to-end claim handling: od zgłoszenia szkody do błyskawicznej decyzji
18:05 - 18:20
Networking Zone
Ask a Speaker: When Data can Act: Security Data Risks in LLM Systems.
18:05 - 18:20
Networking Zone
Ask a Speaker: From Search to Booking: Building Trusted Android Workflows with App Functions
19:00 - 00:00
Kandela Bar, Elektrownia Powiśle
19:30 - 20:30
Kandela Bar, Elektrownia Powiśle
08:00 - 08:10
Online VoD
Generative UI: These Slides Don't Exist Yet
08:10 - 08:20
Online VoD
Jak zbudować agentyczny zespół analityczno-architektoniczny
08:20 - 08:30
Online VoD
Your Agent Has 50 Tools. Now What?
08:30 - 08:40
Online VoD
Your Pipeline Runs in 4 Minutes. The Business Waited 3 Weeks: Measuring Flow, Not Uptime, in Data & ML Platforms
08:40 - 08:50
Online VoD
The Judge Problem: Scaling Production Agents with Dynamic Multi-Model Verification
08:50 - 09:00
Online VoD
AI needs DevOps
09:00 - 09:10
Online VoD
LLMs as Prediction Models: How Prompts Bound the Space of Possible Answers
09:10 - 09:20
Online VoD
Jak NIE robić projektów w Data Science
09:20 - 09:30
Online VoD
Skalowalny web scraping w erze LLM: kiedy AI faktycznie pomaga, a kiedy tylko przepala budżet
09:30 - 09:40
Online VoD
Automating CI/CD Pipeline Remediation: Classifying and Triaging Flaky Tests via LLM Log Parsing
09:40 - 09:50
Online VoD
AI-Driven Test Automation with Claude Code: Real-World Lessons from Production
09:50 - 10:00
Online VoD
AI nie naprawi waszego procesu. Za to zrobi go szybszym
10:00 - 10:10
Online VoD
Scraping Without the Bill: Architecting Resilient Agentic Web Pipelines with Local Language Models
10:10 - 10:20
Online VoD
Harness Engineering - Principles and Practices
10:20 - 10:30
Online VoD
When a Dash Means Zero: Giving AI Trustworthy Context from Open Government Data
10:40 - 10:50
Online VoD
From Model to Platform Product: Engineering Scalable, Multi-Tenant ML Architecture for Cross-Team Deployment
10:50 - 11:00
Online VoD
Before the Pull Request: Detect Multi-Agent Rework While It Is Still Happening
11:00 - 11:10
Online VoD
Beyond Agents: A New Architecture for Reproducible Code Generation Workflows
The event is being held concurrently with Ya!vaConf and Python Summit. With the Multipass package, you get access to all the speeches of the 3 conferences.
For years we create events and propagate knowledge of AI / ML area - now we have prepared a special agenda showing how developers can implement AI solutions even without knowledge of machine learning.
All people involved in the development, implementation and maintenance of data solutions will find a lot of interesting things here
It's great that the agenda is divided into as many as 3 formats - online live, stationary and as a VoD library. It's best to be on-site, but regardless of how and when you prefer to watch, you're sure to learn a lot here.
Online
Monday, October 19th, 2026
available only for registered users.
Onsite
Tuesday, October 20th, 2026
Conference Centre
Wybrzeże Kościuszkowskie 20
00-390 Warsaw
At the event you will meet employers and providers of IT / data science products and services who will introduce you to their company profiles and recruitment offers.
Main organizer
Since 2005, we have been supporting the Polish IT / data science community in its development. We are not only the organizer, but also the initiator of many significant events aimed at industry professionals and students, such as Warsaw IT Days, Data Science Summit, The Hack Summit, Ya!vaConf, MS Tech Summit, Warsaw IT Days Expo and many meetups. We also help organize IT and data science conferences for companies / organizations interested in such activities.