⬇ Download resume (PDF)

PhD candidate in Computer and Information Science at the University of Pennsylvania working at the intersection of applied cryptography, zero-knowledge proofs, and formal methods. Builds and performance-optimizes fast, succinct non-interactive proof systems for formal languages, implemented in Rust and released open-source, with publications at IEEE S&P, USENIX Security, and SOSP. Pairs deep cryptography with three years of security engineering at Meta and a rigorous eye for soundness and correctness. Current focus: using cryptographic tools, including zero-knowledge proofs, to make AI systems auditable and accountable.

Zero-knowledge proofs / SNARKs; verifiable computation; ZK circuit design & prover performance optimization; protocol design & implementation; MPC, threshold signatures, distributed key generation, secret sharing (protocol-level); signature schemes (Schnorr, EdDSA, ECDSA, BIP-340).

Interactive theorem proving (Coq); SMT solvers; program analysis; circuit/constraint compilation (CirC); reasoning about soundness & correctness.

Differential privacy; federated / private analytics; threat modeling, threat & abuse detection, security-sensitive code & design review.

Rust, C++, Python; Coq; LLVM / MLIR; distributed & privacy-preserving systems.

• Build and performance-optimize fast, succinct non-interactive ZK proof systems for formal languages — regular expressions (Reef) and context-free grammars (Coral) — designing ZK circuits and constraint systems and tuning prover runtime and proof size; implemented in Rust and open-sourced.
• Reason rigorously about the soundness, correctness, and security assumptions of cryptographic proof systems using SMT solvers and interactive theorem proving (Coq).
• Apply cryptographic tools, including zero-knowledge proofs, to bring auditability and accountability to AI systems; advised by Sebastian Angel (dissertation: Zero Knowledge Proofs of Formal Languages).

• Designed and implemented zero-knowledge protocols for privacy-preserving fraud detection, taking research designs through to deployable implementations.
• Reviewed and hardened cryptographic tooling to interoperate with modern TLS stacks, bridging research prototypes and production.

• Performed threat modeling and security analysis to design and own company-wide insider-threat and abuse-detection tooling for security and legal investigations teams.
• Scaled alert-processing pipelines through automation, increasing throughput while driving down false-positive rates.
• Led cross-functional security and privacy reviews with Legal and Policy to reduce the data footprint of detection systems without sacrificing efficacy.

• Quantified tradeoffs between differential-privacy guarantees and Voting Rights Act compliance for the 2020 U.S. Census, in collaboration with the U.S. Census Bureau.

S. Angel, S. Celi, E. Margolin*, P. Mishra, M. Sander, J. Woods. [code]

S. Angel, E. Ioannidis, E. Margolin*, S. Setty, J. Woods. [code]

E. Margolin, K. Newatia, E. Roth, T. Luo, A. Haeberlen. * Authors listed in alphabetical order.

Competed at the 2025 World Rowing Championships, Shanghai.