Qiang Zhang

Currently I am a staff AI researcher at Figure AI.Previously, I was a member of technical staff at xAI, working on multimodal generation research and development.Previously, I was a senior applied scientist and tech lead at Adobe Firefly, working on GenAI staffI am an expert in those following: 1. Diffusion Models and Distillation. Proficient in diffusion models such as text-to-image, structure control, style control, image inpainting, text-to-3D, text/image-to-video, video extension, and autoregressive video generation. Skilled in distillation techniques, including DMD, consistency models, GAN loss optimization, ODE regression and so on. 2. Large-Scale Distributed Training. Expertise in FSDP, Tensor Parallelism (e.g., Megatron-LM-TP), Sequence Parallelism (e.g., Ulysses-SP, Ring-Attention, Distributed Key-Value), advanced profiling, checkpointing, mixed-precision training, and optimized attention mechanisms like Flex-Attention. 3. Large-Scale Distributed Data Pipelines. Comprehensive experience in dataset parsing, filtering, preprocessing, and high-performance data loading using SQS, S3, multithreading, multiprocessing, queues, shared memory, Ray, Redis, and FFMPEG.Previously, I was a PhD candidate at Princeton University. I have strong research experiences with first-author top-conference publications in these fields:1. Computer Vision: CVPR 2019 Oral, ECCV 20202. Deep Learning: ICLR 2021 Oral3. Computer Graphics: SIGGRAPH(Asia) 2022 & 20234. Robotics: IROS 2021