Course Syllabus
Learning Objectives
By the end of this course, students will be able to:
- Analyze AI agent methodologies for computer systems design across different abstraction levels
- Evaluate AI-driven design approaches and compare them to traditional heuristic methods
- Synthesize research findings through comprehensive literature review and survey writing
- Identify promising research directions and gaps in AI-agent-driven architecture design
- Communicate complex technical concepts through presentations and collaborative research
Course Structure
This is a research paper reading seminar exploring AI agents across three thematic areas:
- AI for Software: Software optimization and compilation
- AI for Architecture: Hardware design and system architecture
- AI for EDA: Physical implementation and chip design
See the Schedule for detailed weekly topics and readings.
Assignments & Grading
Survey Paper Checkpoints (45%)
- Three milestone checkpoints (15% each): Literature Survey, Data-Driven Exercise, First Draft
- Progressive development of publication-quality research
- Individual work within collaborative framework
Paper Presentations (20%)
- Present and lead discussion for assigned papers
- 20-minute presentation + 20-minute facilitated discussion
- Sign up by Week 3
In-Class Discussion & Attendance (20%)
- Class attendance: Regular attendance at all sessions
- Discussion engagement: Thoughtful contributions to paper discussions
- Peer interaction: Constructive feedback and questions during presentations
Final Paper Integration (15%)
- Collaborative review and finalization of complete survey paper
- Cross-section coordination and quality assurance
- Final presentation of integrated work
See Assignments for detailed specifications and requirements.
Course Policies
Attendance
Regular attendance essential for discussion-based seminar. More than two unexcused absences may result in grade reduction.
Late Work
- Survey paper checkpoints: 10% penalty per day late
- Survey paper project: Extensions require advance notice from entire team
- Discussion leadership: Must reschedule in advance
Collaboration
- Survey paper checkpoints: Individual work within collaborative project framework
- Survey paper project: Required group work (entire class collaboration)
- Discussion leadership: Pairs allowed and encouraged
Generative AI Policy
Philosophy: In a course dedicated to AI agents transforming computer systems design, we embrace the thoughtful use of generative AI tools while prioritizing fundamental learning.
What GenAI Is Good At:
- Explaining concepts: Breaking down complex technical terminology or mathematical notation
- Writing support: Editing, structuring arguments, or improving clarity of your own ideas
- Initial brainstorming: Generating starting points for literature searches or research questions
- Code examples: Understanding implementation concepts through AI-generated snippets
What GenAI Struggles With:
- Research methodology: AI summaries miss the nuanced “how” and “why” behind authors’ methodological choices—this is the gold you need to learn, and I will help you learn that despite having access to ChatGPT and so forth.
- Critical evaluation: AI can’t assess the quality, limitations, or significance of research approaches
- Novel connections: AI can’t make the creative leaps between papers that lead to breakthrough insights
- Domain expertise: AI lacks the deep systems knowledge to evaluate feasibility or identify subtle flaws
How to Use GenAI Effectively:
- Start with your own reading: Engage with papers directly first, then use AI to clarify confusing parts
- Ask specific questions: Instead of “summarize this paper,” ask “what does equation 3 mean?” or “why did they choose this evaluation metric?”
- Verify everything: Cross-check AI explanations against the original source and your growing understanding
- Focus on methodology: Pay special attention to how authors approached problems—AI can’t teach you this critical skill
Remember: Research summaries ≠ understanding research methodology. The methodology is the gold, and as your instructor, I’ll help you develop this skill. Don’t lean too heavily on AI for this—be aware of its limitations.
Learning-First Approach:
- Develop fundamentals first: Before using AI tools, engage deeply with the material yourself. AI should amplify your understanding, not replace it
- Transparent attribution: When AI significantly contributes to your work, acknowledge it (e.g., “Used GPT-4 to help structure the argument in Section 2.3”)
- Critical evaluation: Always verify AI-generated content against primary sources and your own understanding
Why This Matters: The goal isn’t productivity—it’s learning. AI tools are powerful accelerators, but they work best when you have strong foundational knowledge to guide and evaluate their output. In our seminar discussions, we want to hear your insights, questions, and connections, not AI-generated responses.
In-Class Discussions: Our seminar format prioritizes human dialogue and critical thinking. Come prepared to engage with your own questions, insights, and perspectives developed through direct engagement with the material.
Academic Integrity
All work must be original with proper citation. Collaboration encouraged within the framework specified above.
This syllabus may be adjusted based on emerging research and student interests. Changes will be announced in advance.