Miguel de Vega, Ravital Solomon, Remi Gai & Pascal Paillier I Pragma Denver 2024

on YouTube ↗41m

" summary

Panel lays out concrete claims and measurements: compiler tuning produces a 100x performance gap between experts and nonexperts, and a demo workflow runs a sealed-bid auction: 64 bids <1s on a 64‑core machine with simple comparisons around 30 ms. Tools discussed include Concrete Python, EVM‑compatible FHE integrations for onchain confidentiality, and hybrid designs that combine FHE+ZK to offload work. Practical builds include private payments, confidential voting, dark pools and an onchain Mafia game prototype; full encrypted AI at LLM scale still needs GPUs/FPGA.

AI-written summary · about

" chapters (23)

1. 0:00Intro to Privacy Enhancing Technologies Panel
2. 1:26Panelist introductions and backgrounds
3. 3:11What is homomorphic encryption and FHE?
4. 6:14From theory to real-world FHE applications
5. 8:18MPC vs FHE for private computation
6. 10:24Inco’s approach to FHE on EVM-compatible blockchains
7. 13:02Use cases: private voting, auctions, DeFi, gaming
8. 14:09AI, privacy, and encrypted personal data risks
9. 15:48Practical private payment systems onchain
10. 17:12Gaming and confidential mechanics with FHE
11. 18:47Performance benchmarks for FHE today
12. 21:50Circuit optimization and FHE hardware needs
13. 23:47Reinventing algorithm design for private computation
14. 25:21Is hardware dependency a blocker for PETs?
15. 26:22Combining FHE and ZK for better performance
16. 27:00Other PETs: ZK, anonymous signatures, whitebox crypto
17. 29:39Rethinking developer experience for PETs
18. 31:08Compiler challenges for encrypted inputs
19. 32:49Writing performant, privacy-respecting programs
20. 34:41Secure branching and control flow with encrypted data
21. 36:29Concrete Python and automatic FHE compilation
22. 37:35What’s next: private AI, privacy in DeFi, hackathons
23. 40:16Closing thoughts and upcoming announcements

" related talks

• 30mPractical Zero Knowledge ProofsRemi Gai, Kautuk Kundan, Danilo Kim · Pragma India 2023
• 26mWrap-Up Session Panel Discussion: Quick TakesErica Brescia, Anand Kashyap, Marvin Tong · Confidential Computing Summit 2026
• 11mIntroductionsPaul O’Neil · Confidential Computing Mini Summit at Open Source Summit Europe 2023
• 44mGPU-less, Trust-less, Limit-less: Reimagining the Confidential AI CloudMike Bursell · AI Engineer World's Fair 2025
• 17mPreview: Security and PrivacyChia-che Tsai · OSDI '21 / USENIX ATC '21
• 04mRecap VideoStani Kulechov, RENE REINSBERG, Misha Putiatin, Johann Eid, Jordi Baylina, Jan Gorzny, Charles Guillemet, Debbie Soon, Shady El Damaty, Adrienne Youngman, Jonathan Han, Andrej Bencic, E.G. Galano, Josef Jelacic · Pragma Cannes 2026
• 37mFrontiers in CryptographyTony Arcieri · Strange Loop 2016
• 26mAmazingly Fast and Incredibly Scalable Inference...Harry Kim, Laikh Tewari · PyTorch Conference 2025