Applied AI Agents: Using Agents for Work and Research
Course ID 15678
Description AI Agents can work on a variety of digital tasks and increase productive in everyday work and research. In this course, you will learn common patterns of AI agent use that increases productivity on such tasks while mitigating common failure patterns that affect reliability. The course emphasizes working on tasks that would be not possible without agents: handling many parallel tasks; working on tasks and interpreting solutions that one does not understand well; working on a breadth of tasks in different areas at once. Throughout the course you will build your own infrastructure to make your work with agents increasingly productive, reliable, and enable a wide range of tasks.
Key Topics
Basics of AI agents
Agent infrastructure and agent harnesses
Working with claw-style agents
Cognitive and emotional management of agent work
Patterns to work on many tasks at once; tasks that require deep domain expertise
Patterns to work with uncertainty and how to assess the reliability and quality of agent
outputs
Anti-patterns of agent work
Tools and workflows to work with agents in different modalities: text, code, sound, vision, 3D
How to learn while using agents
Required Background Knowledge
Basic research skills; software engineering and software design skills; strong critical thinking skills; basic but broad knowledge of computer science
Course Relevance
The course will require some basic understanding of how components work that agents use: REST APIs, basic security, basic research skills, software engineering and software design. The course will require students to rapidly learn new information that they are not familiar with, which demands some kind of broader knowledge that students acquire over time.
The course might be particularly suitable to master's students. It will contain parts about doing research with agents, which has a great with for PhD students.
Course Goals
Broad: how to be productive with agents while working on everyday tasks and PhD-level research
How to use multiple agents at once to solve multiple tasks
How to use agents to solve complex tasks requiring domain expertise
How to assess if agent-produced outputs are reliable and high quality
How to use agents if problems and outputs are not well understood
Learning Resources
Software is chosen by the students as the course progresses to build their personalized agent infrastructure. Learning resources will mostly be the lectures and the assignments which are centered around learning by doing
Assessment Structure
Autograded homework, 0-100 points (or letter equivalent): 40%
Projects, 0-100 points: 30%
Peer project grading, 0-100 points: 20%
Participation, 0-100 points: 10%
Extra Time Commitment
No extra time commitments.