Berlin DataTalksClub Group

​This workshop updates Module 3: Test, Containerize, and Deploy an AI-Assisted App.

​The previous workshop ended with an application that runs on your machine. In this hands-on session, Alexey Grigorev will take it the rest of the way: proven by tests that exercise the real stack, packaged in containers, checked automatically on every pull request, and deployed at a public URL.

​Each of those steps is something AI tools are good at drafting and bad at owning. A generated `Dockerfile` that builds isn't the same as one that builds the right thing. A green pipeline that skips the tests is worse than no pipeline at all.

## ​What You'll Learn

• ​How to write integration tests that exercise the API, the database, and the frontend together
• ​How to set up and tear down a test database, and when to recreate it instead of rolling back
• ​How to test against the OpenAPI spec instead of against the implementation
• ​How to check that an AI-generated test suite actually catches anything, by breaking the code on purpose
• ​How to containerize the app with a multi-stage build that compiles the frontend and serves it from the backend image
• ​How to move from SQLite to Postgres without rewriting the application code
• ​How to keep secrets and environment configuration out of the image
• ​How to set up CI so linting, tests, and the container build run on every pull request
• ​How to read a failing pipeline log and feed it back to the agent
• ​How to deploy to a platform such as Render, Fly.io, Railway, or Cloud Run
• ​How to handle managed databases, migrations, health checks, logs, and rollback
• ​How to connect the pipeline to the deployment so merging to the main branch is what ships
• ​Why a deploy that succeeded isn't the same as an app that works

​By the end, you'll know how to take an AI-assisted application from your laptop to a public URL that redeploys automatically when the tests pass.

​Like the other workshops, this will be a live demo with practical tips and time for Q&A.

## ​All Events in This Series

## ​All Events in This Series

• ​How to Work with AI Coding Agents: Spec-Driven Development, Context and Loop Engineering, Workflows
• ​Build and Ship an AI-Assisted Full-Stack App
• ​Test, Containerize, and Deploy an AI-Assisted App
• ​Open-Source AI Tools for Security, Audit, and DevOps
• ​AI Dev Tools Zoomcamp 2026 Pre-Course Live Q&A
• ​AI Dev Tools Zoomcamp 2026 Course Launch

## ​​Thinking About Joining AI Dev Tools Zoomcamp?

​This workshop covers the updated content for Module 2 of AI Dev Tools Zoomcamp, our free course on using AI tools in practical software development workflows.

​The course covers AI-assisted planning, coding, review, testing, coding agent capabilities, MCP, skills, plugins, security, audit, DevOps, and a complete final project.

​The new cohort of AI Dev Tools Zoomcamp starts on August 31, 2026. You can join it by registering here.
​

## ​​About the Speaker

​Alexey Grigorev is the Founder of DataTalks.Club and creator of the Zoomcamp series.

​Alexey is a software and ML engineer with over 10 years in engineering and 6+ years in machine learning. He has deployed large-scale ML systems at companies like OLX Group and Simplaex, authored several technical books, including Machine Learning Bookcamp, and is a Kaggle Master with a 1st place finish in the NIPS'17 Criteo Challenge.

Join our Slack: https://datatalks.club/slack.html

​
​About the speaker:
​Ivan Brigida is the creator of PythonInvest.com and a retail investor. He has worked as an analyst for over 15 years and applies those skills to enhance his own investment strategies and research methods.

​He also runs a free course on Stock Market Analytics, which will launch for its third edition in August 2026.

Join our Slack: https://datatalks.club/slack.html

​This workshop updates Module 4: DevOps and Observability for AI-Built Apps.

​The previous workshop ended with an app that deploys automatically when the tests pass. That gets it online. It doesn't tell you whether it still works.

​In this hands-on session, Alexey Grigorev will build the loop that closes that gap: observe user impact, alert with context, investigate from evidence, authorize a bounded response or escalate, verify recovery, and audit both the code and the response trail.

​We also put a coding agent inside that loop as the first line of support. It collects the same evidence you would, compares it with recent changes, and proposes an action. What it doesn't get is production credentials. A model supplies confidence. Code outside the model enforces permission.

## ​What You'll Learn

• ​Which questions you can't answer about your own deployed app right now, and why a dashboard full of CPU graphs isn't observability
• ​How to instrument one important endpoint end to end with OpenTelemetry, so a metric, a trace, and a log all describe the same failed request
• ​What must never reach your telemetry: passwords, tokens, request bodies, user names
• ​How to run the pipeline behind a Collector instead of exporting from the app to each backend, with Prometheus for metrics, Loki for logs, Tempo for traces, and Grafana on top
• ​How to write one alert you would want to be woken up for, with a payload that includes the symptom, the dashboard link, the deployed version, and the runbook
• ​Why forty alerts you've learned to ignore are worse than one you trust
• ​How to collect a bounded evidence packet with allowlisted read-only queries before any model is involved
• ​How to run Codex or Claude Code headless as a read-only investigator that returns facts, ranked hypotheses, a proposed action, and its own uncertainty against a fixed schema
• ​Why authorization belongs outside the model, and what an action allowlist and autonomy levels look like in practice
• ​How to run the same incident twice and watch the policy force an escalation the second time, with no change to the agent
• ​How to layer a security audit: a deterministic scanner for evidence, a model for context and abuse paths, a person to decide
• ​How to inventory the responder's own capabilities in a permission table, because a prompt that says "read only" isn't a read-only token
• ​Where PR-Agent, Semgrep MCP, Snyk Agent Scan, K8sGPT, HolmesGPT, LiteLLM, and Ollama fit once the problems have names

​By the end, you'll know how to tell whether your deployed app still works. You'll also know how to put an agent in the incident loop without giving it production access, and how to audit both the code it ships and the agent that responds.

​Like the other workshops, this will be a live demo with practical tips and time for Q&A.

## ​All Events in This Series

• ​How to Work with AI Coding Agents: Spec-Driven Development, Context and Loop Engineering, Workflows
• ​Build and Ship an AI-Assisted Full-Stack App
• ​Test, Containerize, and Deploy an AI-Assisted App
• ​Open-Source AI Tools for Security, Audit, and DevOps
• ​AI Dev Tools Zoomcamp 2026 Pre-Course Live Q&A
• ​AI Dev Tools Zoomcamp 2026 Course Launch

## ​Thinking About Joining AI Dev Tools Zoomcamp?

​This workshop covers the updated content for Module 4 of AI Dev Tools Zoomcamp, our free course on using AI tools in practical software development workflows.

​The course covers AI-assisted planning, coding, review, testing, coding agent capabilities, MCP, skills, plugins, security, audit, DevOps, and a complete final project.

​The new cohort of AI Dev Tools Zoomcamp starts on August 31, 2026. You can join it by registering here.

## ​About the Speaker

​Alexey Grigorev is the Founder of DataTalks.Club and creator of the Zoomcamp series.

​Alexey is a software and ML engineer with over 10 years in engineering and 6+ years in machine learning. He has deployed large-scale ML systems at companies like OLX Group and Simplaex, authored several technical books, including Machine Learning Bookcamp, and is a Kaggle Master with a 1st place finish in the NIPS'17 Criteo Challenge.
​

Join our Slack: https://datatalks.club/slack.html
​