Fall 2025 Course Topic: Engineering of Human-Centered AI Systems
This course balances conceptual understanding and practical implementation of intelligent multimodal AI assistants that can perform knowledge tasks and automate UI interactions. Through research paper discussions and hands-on laboratories, students learn foundational neural architectures, model adaptation techniques (e.g., RAG, prompt engineering, finetuning, transfer learning, contrastive learning, joint attention, and RLHF), multimodal integration (e.g., text, speech, and gesture), and agent-based interaction. Using industry-standard frameworks (PyTorch, LangChain, Hugging Face), students develop both theoretical knowledge and practical skills in engineering AI assistants, with consideration for data curation, safety, and human factors.
- Semester: Fall 2025
- Instructor: Syed Billah
- Who: Students who can use this course as:
- Ph.D. specialization course or M.S. Elective Course
- When: TR 3:05 p.m. - 4:20 p.m.
- Where: E339 Westgate Building
- Instruction Mode: In person
- Credits: 3.0
- LionPATH Class Number: 15106
Upon successful completion of this course, students will be able to:
- Develop multimodal AI systems.
- Explain the mathematical foundations and intuition behind key machine learning concepts including vector operations, dimensionality reduction, and neural architectures.
- Analyze the evolution and architectural differences between classical neural networks and modern architectures.
- Learn different model adaptation techniques including RAG, prompt engineering, finetuning, transfer learning, and RLHF using industry-standard frameworks.
- Collaborate effectively in team projects for developing complex AI systems.
- Stay current with emerging trends and developments in human-centered AI.
- Class participation and Attendance (10%):
- Each student will present a research paper in the class
- Students must engage in paper discussion
- Assignments (40%): Students will complete 10 weekly assignments:
- Each assignment is a mix of paper review and/or coding task
- Project (50%): The project is divided into 5 parts:
- Part 1: RAG Implementation (15%)
- Part 2: Domain-Specific Chatbot (15%)
- Part 3: Image Integration (15%)
- Part 4: Speech Integration (15%)
- Part 5: UI Task Automation (40%)