agentic AI tools, EECS fundamentals, and applied ML projects overlapping at agency Agency with AI

Effective AI use in applied machine-learning projects

  • 6.S950 Agency with AI: Effective AI Use in Applied Machine-Learning Projects
  • Instructor: Shen Shen
  • Level and units: U, 12 units with a 2-2-8 split
  • Prereqs: 6.3900Introduction to Machine LearningIntroduction to the principles and algorithms of machine learning from an optimization perspective. Topics include linear and non-linear models for supervised, unsupervised, and reinforcement learning, with a focus on gradient-based methods and neural-network architectures. Previous experience with algorithms may be helpful., 6.1010Fundamentals of ProgrammingIntroduces fundamental concepts of programming. Designed to develop skills in applying basic methods from programming languages to abstract problems. Topics include programming and Python basics, computational concepts, software engineering, algorithmic techniques, data types, and recursion. Lab component consists of software design, construction, and implementation of design., and 6.1210Introduction to AlgorithmsIntroduction to mathematical modeling of computational problems, as well as common algorithms, algorithmic paradigms, and data structures used to solve these problems. Emphasizes the relationship between algorithms and programming, and introduces basic performance measures and analysis techniques for these problems., strictly enforced
  • Schedule: Tuesday lectures, Thursday open lab hours. Time and room TBD
  • Description: Examines the effective use of agentic AI tools in applied machine-learning work, such as a UROP or MEng project. Develops both fluency with the tools and agency, the judgment about what to delegate, what to verify, and what to keep for oneself to do. Introduces good agentic-AI practices such as spec-driven development, context management, and harness engineering; where and why LLMs fail by construction, via the machinery beyond vanilla transformers; and how to design evaluations to measure those failures. Each idea introduced connects to a long-standing EECS principle, with attention to what is genuinely new. Includes a semester-long project, with AI-assisted reading checked against sources and compiled into a map of the field, a small ML component built from scratch with an agent, outsourced datasets vetted, and artifacts and their judges audited. A personal portfolio of reusable AI workflows accumulates alongside the project. Enrollment may be limited.
Cite these materials

Shen Shen. Agency with AI: Effective AI Use in Applied Machine-Learning Projects. MIT EECS, 2026. https://shenshen.mit.edu/agencyai. CC BY-NC-SA 4.0.

@misc{shen2026agencyai,
  author       = {Shen Shen},
  title        = {Agency with AI: Effective AI Use in Applied Machine-Learning Projects},
  howpublished = {MIT EECS course materials},
  year         = {2026},
  url          = { https://shenshen.mit.edu/agencyai },
  note         = {CC BY-NC-SA 4.0}
}

Page updated August 29, 2026.