Independent research in applied AI and security.
We investigate how software behaves under real operating constraints: what models can do on available hardware, how to measure their performance, and where complex systems fail.
Local inference, quantization, model evaluation, and the trade-offs between quality, memory, latency, and cost. Our focus is practical experimentation and methods that make results easier to interpret and reproduce.
Research spanning smart contracts, operating systems, embedded devices, and vehicle software.
Selected public work:
- Chronomaly on LG webOS - Linux kernel exploitation and persistent root access on consumer hardware.
- Tesla security research - vulnerability research on Model 3/Y infotainment systems, including root access and persistence.
SourceHat began as Solidity Finance in 2020, initially focused on smart-contract audits. It became SourceHat in 2023 as the security practice broadened. Historical reports remain available in the audit archive.
