Hanzhi Liu

Santa Barbara, California

Hanzhi Liu

I’m a Ph.D. candidate in computer science at UC Santa Barbara, advised by Yu Feng.

I work on programming languages, formal verification, and security. Much of my research asks how we can understand and check the behavior of AI agents, blockchain protocols, and zero-knowledge systems.

I also co-founded Nubit in 2023, where we build Bitcoin infrastructure and verification tools.

WritingScholarGitHubLinkedInX

Conference publications

Preprints ↓
  1. NDSS2027To appear
  2. ACM CCS2026To appear
  3. OOPSLA2025
  4. SBC2025

    Tessel: An Optimizing Compiler for Efficient Zero-Knowledge Circuits

    Junrui Liu, Jiaxin Song, Yanning Chen, Hanzhi Liu, Hongbo Wen, Yanju Chen, Yu Feng

    Conference
  5. ACM CCS2024
  6. IEEE S&P2024

Writing & thoughts

All writing →
  1. Nubit releases Thunderbolt for Bitcoin transfers to offline recipients
  2. Reversing three cards illustrates how recursive programs work
  3. Nubit demonstrates Thunderbolt payments with BRC-20 creator Domo
  4. Fried Rice DAO brings more than 500 people together in Las Vegas

Preprints

  1. arXiv2026
  2. arXiv2026
  3. arXiv2026
  4. arXiv2026
  5. arXiv2026
  6. IACR ePrint2024
  7. IACR ePrint2024
Earlier journal publication
  1. J. SJTU (Sci.)2020

Things I help build

AgentFlow synthesizes multi-agent harnesses for vulnerability discovery.

Puffer Code is a Rust agent runtime and CLI. It has an interactive TUI, a WebSocket daemon for desktop shells such as Bobo, and support for connectors, permissions, skills, tools, provider transports, and session storage.

Pomelo lifts Bitcoin Script into a form we can symbolically verify. It grew out of our work on programmable Bitcoin.

Thunderbolt is our implementation of asynchronous off-chain Bitcoin transfers. The 1.0 release includes a Rust implementation, Bitcoin Core integration, and reproducibility tools.

I also work on finding bugs in real systems, from Ethereum serialization with Alma to Bitcoin Script and AI coding agents. I keep a small collection of public reports and upstream fixes.

Teaching

I really enjoy teaching programming languages, formal verification, and blockchain security. My research on Coda, Tabby, and FORAY connects these areas through proofs of circuit correctness, compiler optimization for zero-knowledge proofs, and attack synthesis for DeFi protocols. My TA work includes programming-language instruction and contributions to blockchain assignments on automated market maker pricing and transaction ordering.

TA (Teaching Assistant)

Talks & presentations

2026

2025

2024

A little more

Before UCSB, I studied electrical engineering and computer science at Shanghai Jiao Tong University (2019–2023), advised by Qinxiang Cao. I started my Ph.D. in 2023.

My research has been supported by Ethereum Foundation Academic Grants in 2024 and 2025. At Nubit, we raised $12M by June 2024, including a seed round led by Polychain.

I wrote the Tornado Crash challenge for BlazCTF 2023. The solution is here, if you’re curious.