AI Engineer: Offensive Security Agents

What you'll do: We're hiring an engineer who lives at the intersection of low-level systems and LLM-driven agents — someone who can read a binary and  build the harness that lets an agent do the same. That means wiring emulation (QEMU, Unicorn, Qiling), instrumentation (Frida, DynamoRIO), fuzzing, and exploit primitives into tool interfaces for agents to reason over, as well as the evaluation & benchmarking infrastructure.

What we're looking for: 2–3+ years across systems programming (C/C++, Rust) and ML infrastructure Comfort with binary analysis, memory corruption classes, and modern mitigations (ASLR, CFI, PAC, MTE)Hands-on work with agent harnesses — orchestration, tool-use, eval loops — in production or at benchmark scale Fluency with at least one emulation stack (QEMU, Unicorn, Qiling, PANDA, FirmAE)Strong RE / debugger chops (GDB, IDA or Ghidra) and the patience to use them

Nice to have: Pwn-heavy CTF experience (DEF CON finals, PPP, Dice Gang, Shellphish, Theori, Team Atlanta), public CVEs, conference talks, AIxCC / CGC participation, or kernel / baseband RE.Read this part carefully That list is what the fully-formed version of this role looks like. It's not a filter. If you're sharp, hungry, and funny, apply anyway — even if half the bullets read like a foreign language today. Tech can be learned. Spirit cannot. We've watched the right people pick up binary exploitation from scratch and outship ten-year veterans inside a year. Tell us what you've taught yourself recently and what you'd tear into first here.

Requirements: U.S. citizenship. TS/SCI clearance eligibility (existing preferred).

What we offer: Competitive cash, meaningful early equity, a mission that matters, and hard problems nobody has solved yet.