RFP: Course Development, Online Course in Agentic AI
The Master of Information and Data Science program at the School of Information at UC Berkeley seeks proposals for an online graduate course in Agentic AI.
About the Course
This 14-week master’s-level online course explores the architecture, development, and orchestration of agentic AI systems. Moving beyond static prompts and basic chat interfaces, the course focuses on designing autonomous entities capable of reasoning, using tools, and collaborating to solve complex, open-ended problems. Students will gain practical experience building robust agentic workflows, implementing multi-agent orchestration frameworks (such as LangGraph, AutoGen, or CrewAI), and ensuring the reliability and safety of non-deterministic systems. The course aims to bridge the gap between prompt engineering and sophisticated cognitive architecture in applied AI.
Student Prerequisites
The instructor should assume that students are well-advanced in data science or related fields. Students must be proficient in Python, have basic understanding of LLM fundamentals (including tokenization, context windows, and basic RAG), and be comfortable working with APIs and asynchronous programming. Prior experience with standard ML libraries and vector databases is expected.
Core Topics
The foundational modules should cover the transition from passive LLM applications to active agentic workflows. Key concepts include:
- Cognitive Architectures: Reasoning patterns including chain-of-thought (CoT), tree-of-thoughts, and ReAct (reason + act) loops.
- Agentic AI Protocols: MCP, A2A, and others.
- Memory & State Management: Implementation of short-term (contextual) and long-term (vector/database) memory, and maintaining state in complex, multi-turn interactions.
- Tool Use & Function Calling: Designing robust interfaces for agents to interact with external APIs, databases, and code execution environments.
- Planning & Decomposition: Techniques for breaking down high-level goals into executable sub-tasks.
Multi-Agent Systems & Orchestration
A significant portion of the course should focus on multi-agent orchestration (MAO) and collaborative patterns:
- Orchestration Frameworks: Hands-on development using frameworks like LangGraph, AutoGen, or Semantic Kernel.
- Agent Communication: Defining protocols for agent-to-agent negotiation, hierarchy, and joint problem-solving.
- Controllability & Persistence: Managing “cycles” or infinite loops and ensuring execution persistence in production.
Advanced Implementation & Evaluation
Other topics that should be considered include:
- Agent Evaluation (Eval): Moving beyond RAG metrics to evaluate trajectory, tool-call accuracy, and task completion rates.
- Compound AI Systems: Integrating agents with specialized models (e.g., small local models for routing vs. large frontier models for reasoning).
- Human-in-the-loop (HITL): Designing patterns for human intervention, oversight, and feedback within autonomous loops.
- Security & Safety: Addressing “jailbreaking”in agentic contexts, prompt injection through tools, and sandboxing code execution.
- Deployment at Scale: Monitoring agentic traces, cost optimization, and managing high-latency reasoning steps in production environments.
About the MIDS Program
The Master of Information and Data Science (MIDS) online program prepares students with the data science skills to assume leadership positions and drive innovation in the field.
Deliverables for Accepted Proposal
Instructors of accepted course proposals will be expected to produce a well-designed, reusable Canvas course. Instructors will collaborate closely with an instructional designer and video producer to ensure the course meets established quality standards and fully aligns with defined learning objectives and outcomes. This partnership is integral to creating a high-impact, student-centered online learning experience.
Submission Requirements
Respondents to this RFP must submit a cover letter and course proposal using the form below. The course proposal should contain at minimum a course description, weekly topic breakdown for a 14-week course, brief descriptions of assignments, grading information, and reading list.
Responses will be accepted until selection is complete.
Strong preference will be given to course developers interested in continuing their association with the School of Information by applying to teach the developed course as a lecturer. The separate lecturer application can be found here: https://aprecruit.berkeley.edu/JPF05270.
Compensation
Compensation for course development will be offered via vendor payment from UC Berkeley. To be eligible to receive compensation, the successful proposer will need to register with the UC Berkeley Accounts Payable Vendoring Team and must meet all applicable university requirements. Our expert team will walk you through the process to ensure that your vendor profile is active before work proceeds. This is not a visa opportunity.
The University of California, Berkeley is an Equal Opportunity/Affirmative Action Employer. All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, age, or protected veteran status. For the complete University of California nondiscrimination and affirmative action policy, see: http://policy.ucop.edu/doc/4000376/NondiscrimAffirmAct
Questions
Questions about this call for proposals can be directed to Amanda Gill, MIDS Academic Program Director.
