The future of data science made easy

Year: 2025 · ▶ Watch on YouTube

Yasmeen Ahmad (Managing Director, Data & Analytics)

Switch language → zh

Segments (9)

  • 00:00:05 · Introduction — Yasmeen Ahmad
    • Yasmeen Ahmad is introduced and sets the stage for demonstrating how data science is made easy with BigQuery, Colab, and Vertex AI, powered by Gemini.
  • 00:00:27 · Demo: The Business Problem — Yasmeen Ahmad
    • A scenario is presented for a consumer goods company with high sales but poor cash flow, using a Looker dashboard to visualize the problem.
  • 00:00:45 · Demo: Data Unification — Yasmeen Ahmad
    • Using a natural language prompt, the Data Engineering Agent in BigQuery unifies siloed data from SAP, Salesforce, and Google Ads into a single table.
  • 00:01:22 · Demo: Data Cleaning — Yasmeen Ahmad
    • Gemini-powered recommendations are used to automatically clean and standardize the newly created multimodal data table with a single click.
  • 00:01:50 · Demo: Multimodal Analysis — Yasmeen Ahmad
    • The Data Science Agent extracts information from unstructured PDF invoices and uses Gemini’s world knowledge to categorize buyers into segments.
  • 00:02:48 · Demo: Root Cause Analysis — Yasmeen Ahmad
    • The agent performs an automated root cause analysis, identifying that a new 36-month payment term promotion is the primary driver of the cash flow dip.
  • 00:03:28 · Demo: Forecasting — Yasmeen Ahmad
    • The analysis is exported to a BigQuery notebook where the agent generates code to create a 3-month cash flow forecast, pinpointing wholesalers as the problem segment.
  • 00:04:18 · Demo: Granular Insights with Heatmaps — Yasmeen Ahmad
    • The forecast is further refined to include product categories, generating a heatmap that provides a surgical insight into which products are most affected.
  • 00:05:08 · Conclusion — Yasmeen Ahmad
    • The demo concludes by highlighting how the entire data science workflow, from ingestion to insight, was completed in minutes, making it ‘easy’.

Products Announced (4)

  • 00:00:23 · Gemini in BigQuery (New Integration)
    • Natural language prompts for data engineering and data science tasks. · Multimodal analysis of structured and unstructured data (e.g., PDFs). · Automated data science workflows for root cause analysis and forecasting.
    • Shown as generally available in the demo.
  • 01:02 · Data Engineering Agent (New Feature in BigQuery)
    • Uses natural language to create data pipelines. · Unifies data from multiple siloed sources (SAP, Salesforce, etc.). · Automatically generates and executes data preparation steps.
    • Part of Gemini in BigQuery.
  • 02:10 · Data Science Agent (New Feature in BigQuery)
    • Generates SQL and Python code from natural language. · Performs multimodal analysis, extracting signals from documents like PDFs. · Leverages Gemini’s reasoning for tasks like buyer segmentation.
    • Part of Gemini in BigQuery.
  • 03:31 · BigQuery Colab Composer (New Feature)
    • Seamlessly exports analysis from a low-code UI to a BigQuery notebook. · Allows AI agent to generate and update code directly within the notebook. · Integrates low-code and code-first data science experiences.
    • Shown as part of the BigQuery notebook experience.

Demos (1)

  • 00:00:27 ✓ · End-to-End Cash Flow Analysis with Gemini in BigQuery — Yasmeen Ahmad
    • A complete data science workflow was demonstrated, starting from a business problem (poor cash flow) and ending with a specific, actionable insight. The demo covered data unification, cleaning, multimodal analysis of PDFs, root cause analysis, forecasting, and generating refined visualizations like heatmaps, all driven by natural language prompts to an AI agent.

Notable Quotes (3)

  • 00:00:16 — Yasmeen Ahmad:

    I’m here to show you the future of data science, made easy.

  • 00:01:30 — Yasmeen Ahmad:

    Easy.

  • 00:05:28 — Yasmeen Ahmad:

    And that, my friends, is the future of data science made, say it with me, easy!

Visual Signals

On-screen (7)

  • 00:00:05 · Yasmeen Ahmad, Managing Director, Data & Analytics
    • Introduces the speaker and her role.
  • 00:00:28 · Looker Dashboard: Revenue & Net Cash Flow
    • Visually establishes the business problem: revenue is increasing while net cash flow is flat or declining.
  • 01:07 · Prompt: 'Pull together data for a cash flow analysis...'
    • Shows the natural language prompt used to initiate the data unification process.
  • 02:14 · Prompt: 'Extract buyers and payment information from invoices and group buyers into categories.'
    • Demonstrates a multimodal query to extract information from unstructured PDF documents.
  • 02:51 · Prompt: 'Identify the root causes of cash flow drop from January to March?'
    • Shows the prompt for the automated root cause analysis.
  • 03:13 · Insights card: 'Payment Terms of 36 Months are the biggest contributor...'
    • Displays the AI-generated insight that solves the core problem.
  • 04:42 · Heatmap of Cash Flow Growth Forecast
    • The final visualization that provides a granular, multi-dimensional view of the business problem for surgical decision-making.

Stage (2)

  • 00:00:07 · Yasmeen Ahmad walks onto the stage to applause.
  • 00:00:12 · Speaker arrives at a futuristic, white, fluted podium with two monitors.

Visual demos (6)

  • 00:00:46 · Google Cloud Console UI
    • A tabbed interface for ‘Orders, Sales & Advertising’ with a ‘Data Engineering Agent’ chat window at the bottom.
  • 01:15 · Automated Pipeline Generation
    • A visual graph showing data sources (SAP, Salesforce, Google Ads) being joined into a single data preparation node.
  • 01:22 · Data Cleaning Recommendations
    • A side panel titled ‘Steps’ with AI-generated suggestions for data transformations, which are applied to clean the table.
  • 02:19 · Generated SQL and Visualization
    • The agent generates SQL code using ML.GENERATE to process documents, and then creates a bar chart of the results.
  • 03:31 · Export to Notebook
    • The user clicks ‘Export as notebook’ and the analysis is seamlessly transferred into a BigQuery notebook UI.
  • 03:53 · Time Series Forecast Chart
    • A line chart showing historical and forecasted cash flow growth, broken down by buyer category, with the ‘Wholesale’ line showing a steep decline.

Key Topics

Generative AI · BigQuery · Vertex AI · Gemini · Data Science · Data Engineering · Multimodal AI · Data Analytics · Low-Code Development · No-Code Development · Forecasting · Root Cause Analysis · Data Unification · Business Intelligence

Takeaways

  • Google is deeply integrating Gemini into BigQuery to create a unified, end-to-end data science platform.
  • New ‘Data Engineering’ and ‘Data Science’ agents allow users to perform complex tasks like data unification and multimodal analysis using natural language.
  • The platform can now analyze structured data alongside unstructured data, such as extracting key information from PDF invoices directly within a query.
  • Automated data science workflows, like root cause analysis and time series forecasting, can be triggered with simple prompts, reducing analysis time from months to minutes.
  • The experience is designed to be seamless between low-code UI (Data Canvas) and code-first environments (BigQuery notebooks), with AI assisting in both.
  • The goal is to democratize data science, making sophisticated analysis accessible to more users and enabling faster, data-driven decisions.
  • The entire process is designed to be ‘easy’.