Data agents: Automated and accelerated
Year: 2025 · ▶ Watch on YouTube
Jeff Nelson (Developer Advocate)
Segments (5)
- 00:00:05 · Introduction to Data Agents — Jeff Nelson
- The speaker introduces the challenge of working with raw data and sets up the presentation on how new Data Science Agents can help.
- 00:41:00 · Demo: Building a Sales Forecast — Jeff Nelson
- The demo begins in a BigQuery notebook, showing how to query and prepare sales data using SQL and Python (BigFrames).
- 01:55:00 · Demo: Using the Gemini Data Science Agent — Jeff Nelson
- The speaker uses a natural language prompt to instruct the built-in Gemini agent to perform feature engineering and generate a sales forecast.
- 03:51:00 · Demo: Creating a Data App — Jeff Nelson
- With a single click, the speaker turns the notebook’s visualization into a fully deployed, interactive, and shareable web application.
- 04:54:00 · Announcing More Data Agents — Jeff Nelson
- The speaker announces the launch of specialized agents for data engineering and conversational analytics, expanding the agent ecosystem.
Products Announced (6)
- 00:28 ·
Data Science Agent in BigQuery(New)- Turns raw data into data apps using natural language. · Integrates directly into BigQuery notebooks. · Collaborative chat-based interface.
- Coming soon to BigQuery notebooks.
- 02:27 ·
Serverless Spark engine in BigQuery(New)- Run Spark code directly within BigQuery. · Seamlessly switch between SQL, Spark, and Python. · No infrastructure management required.
- Available now.
- 02:55 ·
TimesFM Foundation Model(New)- Pre-trained on massive time-series datasets for forecasting. · Accessible directly in BigQuery via the
AI.FORECASTfunction. · Generates forecasts with confidence intervals without model training. - Available now in BigQuery.
- Pre-trained on massive time-series datasets for forecasting. · Accessible directly in BigQuery via the
- 03:53 ·
Data App creation from BigQuery Notebooks(New)- One-click creation of shareable web apps from notebook cells. · Automatically packages assets and provides an external link. · Requires no application development experience.
- Available now.
- 04:55 ·
Data Engineering Agent(Preview)- Specialized agent for data engineering tasks.
- In Preview.
- 04:55 ·
Conversational Analytics Agent(Preview)- Specialized agent for data analysts and business users.
- In Preview.
Demos (1)
- 00:41:00 ✓ · Building a Sales Forecast App with Data Agents — Jeff Nelson
- The entire workflow from a raw data table in BigQuery to a fully deployed, interactive forecasting web application was demonstrated, using a combination of SQL, Python, the Gemini Data Science Agent, and a one-click app creation feature.
Notable Quotes (3)
- 00:22:00 — Jeff Nelson:
I know firsthand just how hard it is to turn that raw data into something useful.
- 03:48:00 — Jeff Nelson:
But here’s a secret: I don’t really like building apps.
- 04:32:00 — Jeff Nelson:
Now that’s powerful.
Visual Signals
On-screen (5)
- 00:05:00 ·
Data agents: Automated and accelerated- This is the official title of the presentation.
- 00:11:00 ·
Jeff Nelson Developer Advocate Google Cloud- Identifies the speaker and his role.
- 04:35:00 ·
A flowchart showing: Raw data -> Data science agent -> Data application- Visually summarizes the end-to-end workflow demonstrated.
- 04:55:00 ·
An expanded flowchart showing 'Data engineering agent' and 'Conversational analytics agent' in previ- Announces the new specialized agents and their preview status.
- 05:03:00 ·
Try the Data Science Agent in Colab [QR Code] goo.gle/data-science-agent- Provides a direct call to action for the audience to try the technology.
Stage (2)
- 00:07:00 · Jeff Nelson walks onto the main stage to applause.
- 04:35:00 · The audience applauds after the successful completion of the end-to-end demo.
Visual demos (6)
- 00:42:00 · BigQuery Notebook UI
- A Colab-style notebook interface inside the Google Cloud console. The speaker pastes and runs SQL and Python code in cells.
- 01:38:00 · Data Visualization in Notebook
- The notebook output switches from a data table to a ‘Chart’ view, displaying a line graph of sales trends and a bar chart of sales by state.
- 01:57:00 · Gemini Data Science Agent Chat
- An ‘Ask Agent’ chat box appears. The speaker pastes a detailed natural language prompt asking the agent to generate a sales forecast.
- 02:05:00 · Agent Code Generation
- The agent thinks and then automatically generates and inserts multiple cells of Python/Spark and SQL code to perform feature engineering and forecasting.
- 03:54:00 · Create Data App Wizard
- The speaker clicks a ‘Create Data app’ button, which opens a modal dialog to select cells and configure the application before deployment.
- 04:15:00 · Deployed Data Application
- A new browser tab opens with a URL, showing the final, interactive web application. The app has dropdown filters and the forecast chart, which can be used by non-technical users.
Key Topics
Data Agents · Generative AI · Google BigQuery · Data Science · Sales Forecasting · Time-Series Analysis · Serverless Spark · Data Applications · Low-Code Development · Natural Language Processing · Gemini · Foundation Models · Data Visualization
Takeaways
- Google is launching ‘Data Agents’ powered by Gemini to automate complex data workflows directly within BigQuery.
- The new Data Science Agent uses natural language to generate code for data preparation, analysis, and model building, drastically reducing manual effort.
- BigQuery now features a serverless Spark engine, allowing seamless use of SQL, Spark, and Python in a unified environment.
- A new foundation model, TimesFM, is now available in BigQuery for high-quality, pre-trained time-series forecasting without needing to train a model.
- Users can convert notebook visualizations into shareable, interactive data applications with a single click, democratizing access to data insights.
- The entire process from raw data to a deployed application is streamlined into a single, integrated workflow within the Google Cloud ecosystem.
- Google is expanding its agent offerings with specialized agents for Data Engineering and Conversational Analytics, currently in preview.