I do research in cryptography, machine learning, and AI security, and I build the systems and open-source tools that go with it. I'm drawn to the places where these fields meet: what a deployed model or protocol reveals to an attacker, how AI systems can be defended, and how to make them efficient enough to run anywhere.
I came to research through engineering, so I care about results that hold up in real deployments and code that others can reproduce. I work mainly in Rust, Python, and TypeScript.
| Area | Focus |
|---|---|
| Cryptography & Cryptanalysis | Protocol security, post-quantum cryptography, practical attack and defense analysis |
| Machine Learning Security | Model extraction, information leakage through model APIs, robust defenses |
| AI & Agent Security | Prompt injection, tool-chain attacks, execution isolation, defense in depth |
| Efficient AI Systems | Sparse architectures, efficient inference, ML on consumer hardware |
| Systems & Performance | Concurrency, I/O, profiling, memory efficiency, reliability |
| Paper | Venue | Links |
|---|---|---|
| Recovery Boundaries and Quantization Limits for Low-Rank Neural Adaptations under Restricted Oracles | Indocrypt 2026 · Springer LNCS | Paper · Code |
| ShieldAgent: Defense-in-Depth Security Framework for Agentic AI Tool Use | IEEE CONECCT 2026 | Code |
| Project | About |
|---|---|
| LockKnife | Android security research and forensics toolkit: Python orchestration with Rust-accelerated components |
| USBoverSSH | Rust framework for securely tunneling and sharing USB devices over SSH |
| BloatwareHatao | Rust toolkit for Android package management, debloating, restoration, and device optimization |





