# 36 | Giza Tech: verifiable ML and Agents

In this episode, we spoke with Ravital Solomon, founder and CEO of Sunscreen. Main topics we discussed:- Past, present and future of FHE adoption- Sunscreen’s architecture and design- Sunscreen tech stack and roadmap- What Sunscreen is doing different- Collaboration with Monad- Hiring tipsLinksSunscreen Website: https://sunscreen.tech/Sunscreen X/Twitter: https://x.com/sunscreentechTwitter: ⁠⁠⁠⁠⁠⁠⁠⁠https://x.com/Ingo_zk⁠YouTube: ⁠⁠⁠⁠⁠⁠⁠⁠https://www.youtube.com/@ingo_zk⁠⁠⁠⁠⁠⁠⁠⁠GitHub: ⁠⁠⁠⁠⁠⁠⁠⁠https://github.com/ingonyama-zk⁠⁠⁠⁠⁠⁠⁠⁠LinkedIn: ⁠⁠⁠⁠⁠⁠⁠⁠https://www.linkedin.com/company/ingonyama⁠⁠⁠⁠⁠⁠⁠⁠

In this episode, we spoke with Jules Drean, co-founder of Tinfoil. Main topics we discussed:- Why privacy matters more in the age of cloud AI- Tinfoil’s architecture: inference in secure enclaves with built-in attestation- Performance vs. privacy tradeoffs between FHE, MPC, and TEEs- Verifiability via transparency logs and reproducible builds- A UX-first approach to private AI, integrating Tinfoil with a single line of code- First users: personal AI agents and regulated industries like healthcare and finance- How Tinfoil enables proprietary model protection alongside user privacy- Expansion beyond inference: future plans for fine-tuning and training in enclavesLinksTinfoil's Website: https://tinfoil.sh/Tinfoil's Twitter/X: https://x.com/TinfoilAITwitter: ⁠⁠⁠⁠⁠⁠⁠https://x.com/Ingo_zkYouTube: ⁠⁠⁠⁠⁠⁠⁠https://www.youtube.com/@ingo_zk⁠⁠⁠⁠⁠⁠⁠GitHub: ⁠⁠⁠⁠⁠⁠⁠https://github.com/ingonyama-zk⁠⁠⁠⁠⁠⁠⁠LinkedIn: ⁠⁠⁠⁠⁠⁠⁠https://www.linkedin.com/company/ingonyama⁠⁠⁠⁠⁠⁠⁠

In this episode, we sat down with Jay Prakash, Co-founder & CEO of Silence Labs. Main topics we discussed:- Silent Shard for MPC-based key management- Silent Compute for secure and private multiparty data processing- GTM focus on privacy tech for banks and finance- Selling privacy tech to banks and institutions- Roadmap toward privacy-as-a-service infrastructureLinks Silence Labs Website: https://silencelaboratories.com/Silence Labs Twitter/X: https://x.com/silencelabs_slSilence Labs Products: https://silencelaboratories.com/productsTwitter: ⁠⁠⁠⁠⁠⁠https://twitter.com/Ingo_zk⁠⁠⁠⁠⁠⁠YouTube: ⁠⁠⁠⁠⁠⁠https://www.youtube.com/@ingo_zk⁠⁠⁠⁠⁠⁠GitHub: ⁠⁠⁠⁠⁠⁠https://github.com/ingonyama-zk⁠⁠⁠⁠⁠⁠LinkedIn: ⁠⁠⁠⁠⁠⁠https://www.linkedin.com/company/ingonyama⁠⁠⁠⁠⁠⁠Join us: ⁠⁠⁠⁠⁠⁠https://www.ingonyama.com/career

In this episode, we spoke with Alan Szepieniec, co-founder of Neptune Cash. Main topics we discussed:- The journey from cryptography to peer-to-peer digital money- Neptune design principles, tradeoffs and parameter selection- The road to launch and the war stories post launch- Using ZK-STARKs at L1- Mutator sets- A lightweight smart contract system- The roadmap towards full decentralizationNeptune's Twitter/X: https://x.com/seagodcryptoNeptune Core: https://github.com/Neptune-Crypto/neptune-core/Twitter: ⁠⁠⁠⁠⁠⁠https://twitter.com/Ingo_zk⁠⁠⁠⁠⁠⁠YouTube: ⁠⁠⁠⁠⁠⁠https://www.youtube.com/@ingo_zk⁠⁠⁠⁠⁠⁠GitHub: ⁠⁠⁠⁠⁠⁠https://github.com/ingonyama-zk⁠⁠⁠⁠⁠⁠LinkedIn: ⁠⁠⁠⁠⁠⁠https://www.linkedin.com/company/ingonyama⁠⁠⁠⁠⁠⁠Join us: ⁠⁠⁠⁠⁠⁠https://www.ingonyama.com/career

In this episode, we spoke with Liz Steininger, CEO, and Anna Kaplan, Security Researcher, from Least Authority. Main topics we discussed:- Least Authority’s roots and their connection to Zcash- Intro to security auditing- What makes cryptography audits unique- Least Authority privacy-first products- Emerging trends in applied cryptography- The promise and limits of AI in auditsLinksLeast Authority X/Twitter: https://x.com/LeastAuthorityWebsite: https://leastauthority.com/Twitter: ⁠⁠⁠⁠⁠https://twitter.com/Ingo_zk⁠⁠⁠⁠⁠YouTube: ⁠⁠⁠⁠⁠https://www.youtube.com/@ingo_zk⁠⁠⁠⁠⁠GitHub: ⁠⁠⁠⁠⁠https://github.com/ingonyama-zk⁠⁠⁠⁠⁠LinkedIn: ⁠⁠⁠⁠⁠https://www.linkedin.com/company/ingonyama⁠⁠⁠⁠⁠Join us: ⁠⁠⁠⁠⁠https://www.ingonyama.com/career

In this episode, we spoke with Sylve Chevet, co-founder and CEO of Hyli, about designing a proof-first L1 that centers ZK as its core primitive. Main topics we discussed:- Hyli brand- How to build a blockchain with a lean team- Hyli testnet: operations and goals- Verifiable Apps and Hyli use cases- The future of ZK- ZK benchmarks for client side proving- Hyli roadmap towards mainnetLinksHyli's Website: ⁠https://www.hyli.org/⁠Hyli's Twitter/X: https://x.com/hyli_orgSylve's Twitter/X: https://x.com/sylvechvYouTube: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.youtube.com/@ingo_zk⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠GitHub: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://github.com/ingonyama-zk⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠LinkedIn: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.linkedin.com/company/ingonyama⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Join us: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.ingonyama.com/career⁠⁠⁠⁠⁠⁠⁠

In this episode, we spoke with Raphael Ducasse, lead engineer at Giza, about building LuminAIR—a verifiable machine learning library—and how it powers autonomous DeFi agents through ZKML. Main topics we discussed:- The shift from Cairo-based tooling to building LuminAIR on Luminol, a minimal Rust-based ML library- How LuminAIR optimize models for ZK proving- Giza’s core thesis, roadmap, why they launched a token and  institutional adoption- Verifiable recommender systems and other ZKML use cases- Giza protocol’s role in decentralizing agent execution with Eigenlayer AVS- How ARMA, Giza’s flagship agent, reallocates funds across lending protocols while maintaining custodyLinksGiza's Website: https://www.gizatech.xyz/Giza's Twitter/X: https://x.com/gizatechxyzLuminAIR: https://github.com/gizatechxyz/LuminAIRYouTube: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.youtube.com/@ingo_zk⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠GitHub: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://github.com/ingonyama-zk⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠LinkedIn: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.linkedin.com/company/ingonyama⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Join us: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.ingonyama.com/career⁠⁠⁠⁠⁠⁠

本期嘉宾：Primus 联合创始人 谢翔（Xiang Xie）#SnarkChocolate 最新一期上线！对话 @PrimusLabs 联合创始人谢翔，聊聊「让真实世界数据安全上链」的正确姿势。- 什么是 ZKTLS？它解决什么痛点？ZKTLS方案中attester（盖戳者）的角色与 TEE 必要性- 中文社区为何对 ZKTLS 讨论度偏低？- QuickSilver 框架缘起与10倍性能优势- Primus 的技术路线、产品形态以及应用场景- ZKTLS与ZKVM的结合方案- 密码学与AI的潜在结合以及Primus在这个方向上的技术储备 相关链接Primus 官网：⁠https://primuslabs.xyz/Primus 推特：⁠https://x.com/primus_labsPrimus 开发者中心: https://dev.primuslabs.xyz/Primus 打赏: https://pay.primuslabs.xyz/技术文档：⁠https://docs.primuslabs.xyz/YouTube：⁠https://www.youtube.com/@ingo_zkGitHub：⁠https://github.com/ingonyama-zkLinkedIn：⁠https://www.linkedin.com/company/ingonyama加入我们：⁠https://www.ingonyama.com/career

In this episode, we spoke with Jonathan Passerat-Palmbach, Senior Research Scientist at Flashbots, about tackling MEV challenges through privacy-enhancing technologies. Main topics we discussed:- Flashbots' evolution: from MEV mitigation to building block building infrastructure- The MEV trilemma and why auctions offer a better approach than spam or latency games- The current MEV landscape and the role of PBS and BuilderNet- How Flashbots leverages TEEs for pre-trade privacy and integrity guarantees in block building - Limitations of TEEs and ongoing research to strengthen trust assumptions- Complementing TEEs with cryptography for stronger defense-in-depth architectures- Experimental work on encrypted backrunning using MPC and FHE- The broader Suave vision and decentralized encrypted mempools- Reflections on federated learning, privacy-preserving ML, and why PETs still struggle for adoption in AILinksFlashbots Website: https://www.flashbots.net/FHE Backrunning Paper: https://fc25.ifca.ai/preproceedings/238.pdfBuildernet: https://buildernet.org/Jonathan’s Projects: https://jopasser.at/YouTube: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.youtube.com/@ingo_zk⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠GitHub: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://github.com/ingonyama-zk⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠LinkedIn: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.linkedin.com/company/ingonyama⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Join us: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.ingonyama.com/career⁠⁠⁠⁠⁠⁠

In this episode, we spoke with Bobbin Threadbare, co-founder of Miden, about designing a new blockchain protocol built from the ground up around client-side proving. Main topics we discussed:- The vision behind Miden: building a blockchain assuming ZK is a native primitive- Why EVM and Solidity fall short for true client-side proving- How Miden’s actor model blends accounts and UTXOs for privacy and modularity- The trade-offs in launching as a centralized L2 before full decentralization- Why client-side proving matters — and what it actually looks like on mobile today- Privacy as a design principle: accounts, nodes, and nullifiers in MidenLinksMiden's Website: https://miden.xyz/Miden's Twitter/X: https://x.com/0xMidenBobbin's Twitter/X: https://x.com/bobbinthYouTube: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.youtube.com/@ingo_zk⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠GitHub: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://github.com/ingonyama-zk⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠LinkedIn: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.linkedin.com/company/ingonyama⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Join us: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.ingonyama.com/career⁠⁠⁠⁠⁠⁠