Talks

A growing collection of presentations and teaching activities — full-class paper presentations, conference and course posters, guest lectures, and TA work.

Course presentations

FlashAttention-1

Topics in CS: Vision-Language Foundation Models seminar, Stony Brook University · Fall 2025

Full class presentation of Dao et al.'s FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness in the SBU VLM seminar. Covered tiling, recomputation, and the IO-aware exact-attention algorithm — framed for an audience already familiar with standard attention but new to memory-hierarchy-aware implementations.

Download slides (PDF)

Posters

VLM-SSM Vision Encoders

SUNY AI Symposium 2026 (SUNY-wide, externally reviewed) · April 2026

Poster on the VLM-SSM vision-encoder work.

Glyph-Aware Text Embedding for Chinese LMs

CSE 538 NLP poster session, Stony Brook University · May 2026

End-of-semester poster — see project page.

Adaptive Chess Tutoring

CSE 537 AI poster session, Stony Brook University · May 2025

Course poster session opened by the instructor to the broader CS department — see project page.

Guest lectures

Embedded Deep Neural Network Processing

National Taiwan University of Science and Technology (NTUST) · December 2023

Host: Prof. Sheng-Chang Juan. Ministry of Education Smart Chip Systems and Applications Course Promotion Program, ROC year 112 (2023). 12-hour course taught jointly by three Inventec instructors (Lee Cheng-Lin, Lin Chia-Ching, and me); I taught the December 12 session (3 hours). Material drew on the Inventec AI-on-Chip team's accelerator-engineering experience.

Topics covered:

Model training for resource-limited hardware Model quantization Hardware verification DNN accelerator architecture Algorithm–hardware co-design

Slides not publicly distributed (Inventec internal material). Detailed write-up available on request.