Accelerating data workflows with AI Agents in your IDE
Data practitioners today bounce between notebooks, consoles, CLI tools, and dashboards just to go from idea to deployed pipeline — and every context switch costs time and focus. In this session, we'll demo how Kiro, connected to a cloud data environment via SageMaker Unified Studio's Local IDE feature, collapses that multi-tool, multi-tab reality into a single, prompt-driven experience. You'll watch us go from a raw dataset to a governed, queryable asset — all from one IDE, one conversation —
The Problem: The episode opened with Tony Mullen (Barcelona), new co-host Lakshmi (Düsseldorf, Specialist SA for Data & Analytics), and guest Bhargav (Bay Area, Senior Software Engineer on SageMaker with 6 years' experience). They framed the core challenge facing data practitioners today: the typical data journey — getting access to data, ingesting it, managing permissions, performing quality checks, and running analytics — involves juggling 10+ browser tabs, multiple consoles, and scattered context across different tools, often costing half a day before any real work begins.
The Solution: Bhargav demonstrated how SageMaker Unified Studio combined with AI agents in local IDEs (Kiro, VS Code with Claude Code) collapses this entire workflow. Using two personas — Alex (data engineer) and Sam (data scientist) — he showed how agents can ingest CSV files into a Glue catalog, run automated data quality validation notebooks, perform complex analytics via natural language prompts (no SQL or code required), generate visualizations, and produce executive-level summaries with actionable recommendations — all within a single IDE session connected to cloud compute via SSH.
Key Tooling & Takeaways: The demo showcased several key technologies working together: MCP (Model Context Protocol) servers providing agents with tool access to AWS services, agents.md steering files for providing domain context and guardrails, the AWS Agent Toolkit (open-source bundle of skills, MCPs, and guardrails that works across Claude Code, Kiro, and Codex), and AWS Toolkit for IDE-based exploration of data catalogs. All agent actions are auditable via CloudTrail and governed by IAM policies. The show closed with the announcement that the next episode (1 June) will cover "Using Agentic AI to Power the Data Platform and DataOps."
Key Highlights
- Core demo: End-to-end data workflow — from raw CSVs to executive brief — done entirely via AI agents in ~15 minutes
- SageMaker Unified Studio: Single pane for data, analytics & ML; supports JupyterLab and Code Editor spaces with scalable cloud compute
- IDE integration: Spaces connect to local IDEs (Kiro, VS Code, Cursor) via SSH through AWS Toolkit
- MCP servers: Data Processing MCP provides 36+ tools (Glue, Athena, EMR) that agents can call autonomously
- agents.md / steering files: Standard way to provide domain context, preferences, and guardrails to agents
- AWS Agent Toolkit: Open-source package bundling skills + MCPs + guardrails; works with Claude Code, Kiro, Codex, and more
- Live demo results: 3.4M+ orders processed; agent produced heatmaps, basket-size analysis, and actionable business recommendations (e.g. "launch basket incentive program", "run midweek flash sales")
- Security & governance: All agent actions use space credentials, logged in CloudTrail/CloudWatch, constrained by IAM policies
- Audience poll: Majority of viewers already using Kiro
- Next episode: 1 June — Using Agentic AI to Power the Data Platform and DataOps
Hosts of the show
Tony Mullen - Senior Database Specialist SA @AWS
Lakshmi Nair - Principal Specialist SA @AWS
Guests
Bhargava Varadharajan - Software Engineer @AWS
Links from today's episode
| Topic | Link |
|---|---|
| Community Page | https://lets-talk-about-data-aws.com/past-shows |
| Kiro | https://lets-talk-about-data-aws.com/s415-kiro |
| SageMaker Unified Studio | https://lets-talk-about-data-aws.com/s415-sagemaker |
| Analytics Skills Plugins | https://lets-talk-about-data-aws.com/s415-agent |
| Integration of Kiro and SageMaker | https://lets-talk-about-data-aws.com/s415-kiro-sagemaker |
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article