Coding Agents Meet Data Science

2026-03-26 · Source: The Data Exchange · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Software Development & Engineering · Depth: Intermediate, extended

Summary

Mikio Braun, Senior Principal Applied Scientist at Zalando, discussed the practical application of AI-powered coding agents, particularly within data science workflows. He highlighted that while coding agents, like Claude Code, excel at generating code, they often lack the skepticism and domain knowledge crucial for data science tasks, frequently jumping to conclusions with unvetted data. Braun noted that agents perform better with clean, structured warehouse data and iterative modeling, suggesting a need for data-science-specific agents. The conversation also explored the team-level implications of increased developer velocity, identifying code review, testing, and deployment as new bottlenecks. Braun shared his "vibe-coded" side projects: Talk with Ren, an AI-powered conversational language practice tool for Japanese and other languages, and Bjorn the Bouncer, a text adventure experiment, both built using AI agents.

Key takeaway

For CTOs and VPs of Engineering grappling with integrating AI coding agents, recognize that while individual developer velocity will surge, your team's overall throughput will be constrained by existing code review, testing, and deployment processes. Prioritize investing in AI-assisted tools for these bottleneck areas and foster a culture where developers cultivate "systems thinking" to effectively evaluate AI-generated code, rather than solely focusing on code generation. This shift demands adapting your development infrastructure to support higher velocity and more frequent code changes.

Key insights

AI coding agents boost developer productivity but introduce new bottlenecks and require specialized adaptation for data science workflows.

Principles

• AI agents accelerate coding but lack data skepticism.
• High velocity shifts bottlenecks to review and testing.
• Domain knowledge remains critical for AI output evaluation.

Method

Mikio Braun "vibe-coded" side projects by iteratively building and refining code with AI agents, demonstrating a collaborative, exploratory approach to development rather than strict spec-driven execution.

In practice

• Use AI agents for iterative coding and rapid prototyping.
• Develop systems thinking to evaluate AI-generated code.
• Explore AI for language practice and interactive content creation.

Topics

• Coding Agents
• Data Science Workflows
• AI-Augmented Productivity
• Code Review Bottlenecks
• Systems Thinking

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Scientist, Data Scientist, Software Engineer

Related on AIssential

• AI Coding Assistants Haven’t Sped up Delivery Because Coding Was Never the Bottleneck — AI coding tools shift bottlenecks from coding to specification and verification, demanding new team structures and workflows.
• Shopify/liquid: Performance: 53% faster parse+render, 61% fewer allocations — AI-driven "autoresearch" can yield significant performance gains even in mature, highly optimized codebases.
• Agentic Code Review — AI agents make code writing cheap, shifting the software engineering bottleneck to human verification and review.
• The Mythical Agent-Month — AI agents accelerate code generation but amplify essential complexity and introduce new forms of technical debt.
• Quoting James Shore — AI coding agents must inversely reduce maintenance costs to justify increased code output.
• It’s starting… — AI systems are increasingly automating their own development, accelerating progress while shifting human roles to high-level judgment.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Data Exchange.