PChiang Lab AI Design · Fudan Microelectronics Contact
中 EN
Back to course home

ICSE70011.01 · 2026-2027 Academic Year · Semester 1

AI Design: Introduction & Fundamentals

From Transformers and diffusion models to multimodal agents, inference systems, and AI hardware, build a complete path from principles to applications.

GraduateCurrent course
Credits / hours
2 / 36
Language
Bilingual
School
Fudan University School of Microelectronics
Instructor
Full Professor Patrick Chiang

Course overview

From core models to real applications.

This graduate course introduces the principles and practice of artificial intelligence design and software development. Lectures, hands-on work, and discussion are combined with simplified model implementations, API experiments, and an integrated project.

LevelGraduate
Total hours36 hours
FormatLecture, practice, discussion
PrerequisitesNo specific prerequisite

Learning goals

Move from understanding to design.

  • Understand the principles and development of core models including Transformers and diffusion.
  • Explore inference extensions, BitNet accelerators, and memory-free computing.
  • Implement simplified models, use AI APIs, and design image, video, and multimodal agents.
  • Build interdisciplinary problem-solving and collaboration skills through a complete project.

Course projects

Choose a direction and make a prototype.

Start with a clear scenario, test methods step by step, and use a few metrics to record what changes. The project workspace contains the full directions.

01AI Systems

Model Router + RSL

Build an AI gateway that selects models by task, budget, latency, and risk, then improves through independently evaluated, reversible updates.

02AI for Hardware

Optical Module Tuning

Use machine learning and experimental data to find optical-module settings that satisfy bit-error, power, temperature, and stability constraints.

03Physical AI

LiDAR 3D World Model

Build a traceable pipeline from LiDAR point-cloud collection and synchronization through labeling, quality control, and a 3D world-model evaluation.

04AI Hardware

Hardened-Weights Chip

Map a trained, fixed-weight neural network to dedicated inference hardware and measure the trade-offs in area, latency, throughput, and energy.

05Generative AI

Video Generation on Low-Cost GPUs

Optimize a video-generation or video-to-video workflow for predictable memory, speed, and quality on a single consumer GPU.

06Code Intelligence

Small Coding Model Post-Training

Post-train an open coding model with traceable, permitted data and demonstrate improvement on a held-out programming task without evaluation leakage.

Open the project workspace

Assessment

Complete the course with verifiable results.

20%

Assignments

Model implementations, API experiments, reading notes, and short technical reports.

40%

Project seminar

Present and discuss a group project focused on a real problem.

40%

Project paper

Report the method, results, failures, limitations, and next steps.