Wei Ao

Applied Scientist at Amazon, working on multimodal

[Wei Ao @ao_wei]

#ECCV 2026 is just around the corner! I’ll be heading to Malmö, Sweden, to present our work on long and streaming video understanding, QSVideo. @eccvconf @Lannnn_w @VishnuBoddeti
🗓️ Thursday 4:30–6:30 PM 📍 ExHall #233
📄 Paper: https://t.co/4VxzcPXqb1

[Wei Ao @ao_wei]

@solsol_choi @eccvconf Wow 🤩

[Wei Ao @ao_wei]

Discussion ongoing #ECCV2026

[Wei Ao @ao_wei]

Say Hi #ECCV2026

[Wei Ao @ao_wei]

Next stop Copenhagen #ECCV2026

[Wei Ao @ao_wei]

Arrived at Iceland and enjoyed delicious Elda breakfast. Next stop is Copenhagen. Excited to meet new and old friends at #ECCV2026 Please let me know if you want to grab a coffee or have a quick talk.

[Wei Ao @ao_wei]

Anyone flying to #ECCV from Seattle?

[Wei Ao @ao_wei]

Beyond video understanding, my current research focuses on multimodal and LLMs. I’ve been an Applied Scientist at Amazon Personalization since May. If you’ll be at ECCV, I’d love to connect — whether to talk research, industry, or just grab a coffee.

[Wei Ao @ao_wei]

#ECCV 2026 is just around the corner! I’ll be heading to Malmö, Sweden, to present our work on long and streaming video understanding, QSVideo. @eccvconf @Lannnn_w @VishnuBoddeti
🗓️ Thursday 4:30–6:30 PM 📍 ExHall #233
📄 Paper: https://t.co/4VxzcPXqb1

[Wei Ao @ao_wei]

QSVideo is inspired by how humans understand videos. It models semantic relevance, including Object, Action, and Location. It jumps to likely relevant moments, explores nearby evidence, and searches across time.
🎥 Video: https://t.co/6f1V9xQ1i6

[Wei Ao @ao_wei]

🎯 Excited to share that our paper on long and streaming video understanding has been accepted to #ECCV2026! 
👏 Special kudos go to my co-author @Lannnn_w and my advisor @VishnuBoddeti. Stay tuned for more details!

[Wei Ao @ao_wei]

✨ Features:
1️⃣ PyTorch training code
2️⃣ Microsoft SEAL FHE inference code (C++)
3️⃣ Seamless pipeline bridging PyTorch → SEAL
CryptoFace enables privacy-preserving face recognition without exposing raw images or features. Check it out and ⭐ the repo if you find it useful!

[Wei Ao @ao_wei]

🚀 We have open-sourced CryptoFace (CVPR 2025)!
🔒 The first end-to-end secure face recognition system on encrypted face images with Fully Homomorphic Encryption (FHE).
👉 Repo: https://t.co/IHC8RxJH2T

[Wei Ao @ao_wei]

📢 16:00-18:00 Today! ExHall D Poster #340 @CVPR 2025.
🎯 It is well known that language can help visual search. Can a generated image help visual search? Our answer is YES.

[Wei Ao @ao_wei]

17:00-19:00 Today! ExHall D Poster #324 @CVPR 2025

CryptoFace is the very first end-to-end secure solution for face recognition. We provide the security for both biometric data and the database.

[Wei Ao @ao_wei]

We just finished the tutorial, Computer Vision over Homomorphic Encrypted Data @CVPR 2025.
All slides are available at https://t.co/JQWIBEKECy. We will release the video soon. Thanks to the audience for coming.

[Wei Ao @ao_wei]

My talk is at 10:35. You can find more details at https://t.co/iNPnktOtUb. I will stay at Notre Dame until the afternoon. Welcome to talk to me.

[Wei Ao @ao_wei]

This Saturday, I will give a talk on my CVPR paper at the University of Notre Dame. In pursuing safe artificial intelligence (AI), I will introduce the first end-to-end secure face recognition system enhanced by FHE.

[Wei Ao @ao_wei]

🔉 I have two papers and one tutorial accepted @CVPR Stay tuned! #CVPR2025

🎯 I am actively seeking summer and fall interns. I focus on (1) AI Security enhanced by FHE and SMPC; (2) LLM and multimodal. Please visit: https://t.co/F1Oq745PYA.

DM me if you have any opportunities.

[Wei Ao @ao_wei]

@SepDehdashtian many congratulations!!!

[Wei Ao @ao_wei]

Recently, we open source our solution AutoFHE: https://t.co/SnPjbd0tnu.
We understand the engineering challenges of applying homomorphic encryption to deep learning. Let's work together to make AI and your data safer.

[Wei Ao @ao_wei]

Homomorphic encryption is the answer to your data privacy in the age of deep learning. Apple made it!
https://t.co/x58e3xooKP