Welcome

Welcome, everyone, to AI-Assisted Development: Practices and Pitfalls, a one-day workshop for UBC Master of Data Science students.

AI coding tools are reshaping data science and software engineering. Whether you end up in analytics, ML engineering, or somewhere in between, fluency with these tools and a clear-eyed understanding of their risks are quickly becoming non-negotiable.

This one-day, hands-on workshop covers the practical, ethical, and career dimensions of AI-assisted software development. You will learn effective prompting strategies, best practices for working with agents like Claude Code or Copilot, and techniques for verifying AI-generated code. You will also grapple with the hard questions: privacy risks, vendor dependency, intellectual property, environmental impact, and what these tools mean for the future of technical careers. In small groups, you will set up and run open-weight models locally on your own hardware and then compare their performance against cloud-based tools on a real coding task.

This is not a topic you can figure out later. The landscape is changing faster than anyone can keep up, opinions are sharply divided, and yet proficiency with these tools is quickly becoming an expected skill. The choices you make and habits you form now will have real consequences for your work and career. Frameworks for navigating all of this are currently lacking, so there is fertile ground for exploration and discussion.

Learning Objectives

By the end of this workshop, you will be able to:

  1. Use AI-assisted coding tools effectively to accomplish real software development tasks, including crafting well-structured prompts and managing context
  2. Critically evaluate and verify AI-generated code, identifying common failure modes such as hallucinated APIs, subtle logical errors, and fabricated dependencies
  3. Articulate the distinction between “vibe coding” and responsible AI-assisted development, and explain why it matters
  4. Set up and run open-weight language models locally using tools like LM Studio, and connect them to agentic coding harnesses
  5. Assess the tradeoffs between cloud-based and local AI tools along dimensions including performance, privacy, cost, reliability, and environmental impact
  6. Identify and reason about the risks of AI-assisted development, from prompt injection and data exfiltration to vendor lock-in, IP concerns, and regulatory exposure
  7. Navigate the ethical dimensions of AI-assisted coding, including environmental impact, corporate accountability, deskilling, and societal harms
  8. Describe how the role of the software developer and data scientist is evolving, and articulate a personal strategy for staying effective in a rapidly changing landscape