est. 2026 · colorado · independent

Adversarial AI,
applied.

An independent research and engineering lab at the intersection of AI systems and cybersecurity. Model security, agent hardening, offensive R&D — done quietly, for serious teams.

research → engagements →

// current focus

Research areas under active work.

  • [ AI-RED-TEAM ]

    Adversarial evaluation

    Probing alignment and safety failures in production LLMs, multi-agent systems, and the orchestration layers around them.

  • [ MODEL-SEC ]

    Model security & integrity

    Abliteration analysis, fine-tuning attacks, training-data poisoning, and supply-chain risk in open-weight ecosystems.

  • [ AGENT-SEC ]

    Agentic system hardening

    The new attack surface — tool-use, MCP, autonomous workflows, and prompt injection in production agent stacks.

  • [ OFFENSIVE-R&D ]

    Offensive security R&D

    Traditional pentest research, payload engineering, and lab work on owned hardware. Tools, not theater.

// engineering

For teams shipping AI in production.

Targeted assessments of model security posture, red-team engagements against agentic systems, and consulting on secure AI infrastructure. Discreet, NDA-friendly, senior. No tiered support queues — you talk to the engineer doing the work.

see services → dev@aesirkode.com ↗