ECE605 Large-Scale AI

Course Overview

The Fall 2026 offering focuses on the Embodied AI and Visual-Language-Action models. The course emphasizes real-world research problem-solving through three core pillars:

• Literature Review: Role-playing presentations dissecting frontier across five thematic clusters:

• Final Project Foundations & Interactive Simulation

• Vision-Language-Action (VLA) Architectures

• Robotics Data Systems & Scaling Laws

• World Models & Latent Planning

• Hybrid & Multimodal FrontiersVision-Language-Action (VLA) architectures and generative control policies.

• Rigorous Experimentation: Hands-on empirical exploration, baseline reproduction, and model training on GPU clusters.

• Collaborative Teamwork: Large-team engineering to build, iterate, and submit competitive solutions to a live benchmark.

Class schedule

• Monday and Wednesday, 10:30-11:45, Moore Hall 203 (zoom link; password: ece)

• Office hours: by appointment or after class

Prerequisites

• Essential background: ECE405 Introduction to Large-Scale AI Systems (undergraduate version). Mastery of the material covered in ECE405 (or equivalent) is crucial for success in ECE605. If you have not taken it, please review the ECE405 lecture slides, recordings, and assignments.

Core materials

• Course Syllabus

• Seminar and Project schedule (Sign-up sheet):

  • list your Discord handle and ACCESS username

  • pick your project team

  • indicate interest in reporting on a paper

  • suggest a paper for review

  • find out the schedule of papers to be reviewed

• Paper Seminars Materials folder: shared Google Drive pool for uploading seminar slides and notes (see syllabus for role guidelines and formatting)

• Final project challenge: 2026 BEHAVIOR Challenges (see syllabus for milestone deliverables and rubrics)

• Assignment Submission Form (For Weekly reports and project milestones)

• Discord forum

• Recordings (need to request access)

• Lecture slides

Additional materials

• Koa cluster onboarding guide

• Dive into Deep Learning book

• Deep Learning (MIT) book

• Huggingface daily papers

• Reading list

• Anonymous feedback form