Meet the Lagrange Interns (2024 Cohort)
Each year, Lagrange’s internship program brings together exceptional students and researchers working at the frontier of cryptography and verifiable computation. Under the direction of Charalampos (Babis) Papamanthou—Chief Scientist at Lagrange, Professor of Computer Science at Yale, and Co-Director of the Yale Applied Cryptography Laboratory—interns explore both foundational research and practical systems design for the emerging infrastructure of verifiable AI and secure computation.
Papamanthou’s mentorship combines academic rigor with real-world systems engineering. His research in verifiable computation, privacy-preserving systems, and blockchain security has shaped much of the modern cryptographic landscape, with over 15,000 citations and support from the NSF, JP Morgan, and Protocol Labs.
The 2024 cohort advanced that legacy through projects spanning fair zero-knowledge marketplaces, dynamic proof systems, and streaming SNARKs—each contributing to the broader goal of making cryptographic proofs more efficient, adaptive, and deployable in real-world systems.
Arthur Lazzaretti
Yale University
Arthur’s project explored how to design a fair marketplace for zero-knowledge proofs, where clients (bidders) request computation and provers (sellers) offer distributed compute power. He formalized the problem as a two-sided market and analyzed it from both cryptographic and economic perspectives.
The result was a double-auction mechanism that satisfies key economic and security properties, including truthfulness, weak group-strategy-proofness, weak budget balance, and computational efficiency. The mechanism achieves a strong approximation of maximum social welfare and can generalize to other resource allocation problems where participants trade computational capacity. Arthur’s research lays the groundwork for decentralized, economically fair proof markets, a step toward scalable zero-knowledge systems that balance incentives with verifiability.
Weijie Wang
Yale University
Weijie’s work introduced a new class of proof systems called dynamic zk-SNARKs—zk-SNARKs that can be updated sublinearly as the underlying data or statement changes, rather than recomputed from scratch. In this model, an update algorithm transforms an existing valid proof into a new one with significantly lower computation cost when the new input differs only slightly from the original. Weijie presented two concrete constructions—Dynaverse and Dynalog—that achieve sublinear update time and compact proof sizes.
These systems enable new applications such as sparse zk-SNARKs, where proving complexity depends only on nonzero witness components, and bounded incremental verifiable computation (BIVC), where proofs evolve efficiently over state changes. Weijie’s work establishes the foundation for adaptive, stateful proofs—a key requirement for real-time verifiable computation and dynamic AI systems.
Hadas Zeilberger
Yale University (PhD Candidate)
Hadas’s research focused on building a streaming SNARK—a proof system that generates succinct non-interactive proofs while reading large witnesses directly from disk, rather than loading them entirely into memory.
While modern SNARKs can prove large statements efficiently, they remain constrained by high RAM requirements. Hadas’s work introduces a streaming approach that decouples proof generation from in-memory data storage, enabling low-memory devices to participate in large-scale proof generation. This innovation expands the accessibility of zero-knowledge proofs to commodity hardware and edge devices, opening the door for verifiable computation to run securely on laptops, IoT nodes, and embedded systems.
