# Ilan Komargodski

> Ilan Komargodski is a cryptographer and computer science professor known for research in cryptography and for co-founding Pearl Research Labs.

- Canonical URL: https://iq.wiki/wiki/ilan-komargodski
- Categories: People in crypto
- Tags: Founder, Researcher
- Created: 2026-10-18T06:15:51.077Z
- Last updated: 2026-10-18T06:15:51.077Z
- Source: IQ.wiki — the world's largest blockchain and crypto encyclopedia (https://iq.wiki)

---

**Ilan Komargodski** is a cryptographer and associate professor in the School of Computer Science and Engineering at the Hebrew University of Jerusalem. His research focuses on the foundations of computer science, particularly cryptography. He is also a co-founder of Pearl Research Labs, where his public work has included privacy-preserving proofs for artificial-intelligence computations.[\[1\]](#cite-id-rtqx3ulrjb)​[\[2\]](#cite-id-2bvsfhh26a)​

## Education and Career

Komargodski earned his Ph.D. at the Weizmann Institute of Science under the supervision of Moni Naor and his M.Sc. there under Ran Raz. He later worked as a postdoctoral researcher at Cornell Tech, hosted by Rafael Pass and Elaine Shi, before joining the Hebrew University, where he is an associate professor.[\[1\]](#cite-id-rtqx3ulrjb)​

At the Hebrew University, Komargodski has advised graduate students and worked with postdoctoral researchers and interns. Those listed in connection with his group include postdoctoral researcher Ohad Klein and doctoral student Yuval Gelles, as well as master’s students Youlong Ding, Shay Gadot, and Reut Oelbaum. His listed interns include Cody Freitag, Ashrujit Ghoshal, Ke Wu, Wei-Kai Lin, Aayush Jain, and Rex Fernando.[\[1\]](#cite-id-rtqx3ulrjb)​

### Complexity and Cryptographic Foundations

Early work with Ran Raz examined lower bounds for computational formulas. Their papers included “Average-Case Lower Bounds for Formula Size” at the 2013 Symposium on Theory of Computing (STOC) and “Improved Average-Case Lower Bounds for DeMorgan Formula Size” at the 2013 IEEE Symposium on [Foundations](https://iq.wiki/wiki/foundation) of Computer Science (FOCS), later published in _SIAM Journal on Computing_ in 2017. Other work in complexity and cryptographic foundations included “Hardness Preserving Reductions via Cuckoo Hashing,” with Itay Berman, Iftach Haitner, and Moni Naor, and “White-Box vs. Black-Box Complexity of Search Problems: Ramsey and Graph Property Testing,” with Naor and Eylon Yogev.[\[1\]](#cite-id-rtqx3ulrjb)​

### Secure Computation and Oblivious Memory

A substantial part of Komargodski’s research concerns secure computation and oblivious random-access memory (ORAM), a method for accessing data while concealing the access pattern. His work includes “OptORAMa: Optimal Oblivious RAM,” with Gilad Asharov, Wei-Kai Lin, Kartik Nayak, Enoch Peserico, and Elaine Shi, and “Oblivious RAM with Worst-Case Logarithmic Overhead,” with Asharov, Lin, and Shi. The former appeared at EUROCRYPT in 2020 and was later published in the _Journal of the ACM_; the latter appeared at CRYPTO in 2021 and was later published in the _Journal of Cryptology_.[\[1\]](#cite-id-rtqx3ulrjb)​

## Pearl Research Labs

Komargodski describes himself as a co-founder of Pearl Research Labs. His posts discuss the use of zero-knowledge methods to verify artificial-intelligence inference: such proofs could allow AI laboratories to demonstrate that computations followed specified rules without revealing model weights or prompts. He has characterized zero-knowledge machine learning as an underappreciated part of the AI pipeline and praised work by the Attestable team.[\[2\]](#cite-id-2bvsfhh26a)​

He has also discussed the computational design of Pearl Protocol. In a post describing the project’s then-current approach, he said its matrix multiplication used integers and therefore required model quantization. He described development of a next-generation [Proof of Useful Work](https://iq.wiki/wiki/proof-of-useful-work) intended to support floating-point computations natively and remove that quantization overhead. In another post, he discussed the implementation of Mixture-of-Experts workloads, contrasting expert-by-expert matrix multiplication with alternative ways to apply each expert to its assigned tokens.[\[2\]](#cite-id-2bvsfhh26a)​