Dear Anthropic Hiring Team,

I am applying for the Research Engineer, Agents position on the Agentic Systems team. I am Director of Operations at Percival Engineering and, in parallel, Senior Cybersecurity Systems Engineer / Agentic AI Systems Engineer at the ARKONA Ecosystem (Odenton, MD), where since 2025 I have architected and operated a production autonomous platform with 55 services and 18 AI agents across 6 domains — built entirely through pair programming with Claude Code. It is a running system that manages itself around the clock, not a demo.

Three components of ARKONA are directly relevant to this role:

COMET — AI governance framework. Decomposes organizational workflows into tasks and classifies each with a 5-level delegation model and RACI matrix, grounded in NIST AI RMF, ISO 42001, and IEEE standards. COMET decides what agents should own, what requires human oversight, and what stays manual.

Skill Builder pipeline. Transforms COMET's governance output into production-ready local agents. Uses Anthropic's Agent SDK to construct task-specific agents from RACI classifications; those agents run on Claude, logging every input/output pair as training data. Once 1,000+ validated examples accumulate, QLoRA fine-tuning produces specialized local models on dual NVIDIA P40 GPUs. Graduated models route through MuXD, our hybrid LLM router, which has cut cloud API cost by 47% by deciding local-vs-cloud per request.

Full agent infrastructure. Structured communication, persistent memory, context compression, circuit breakers, thermal management, and autonomous battle rhythm — research agents at 02:00, night builds at 03:00, daily summaries at 06:35 — without human intervention.

Before ARKONA, I spent 20+ years as a cybersecurity systems engineer across the U.S. Air Force, NSA, and the defense industry. NSA Computer Network Operations Development Program graduate. Engineered and led weapon-system cybersecurity programs with NSA, ACC, AFSOC, AFTC, AFLCMC, and PACAF — covering aircraft avionics, acquisitions, systems engineering, test and evaluation, system threat analysis, and vulnerability assessments. That background in safety-critical systems shapes how I approach agent design: failure modes first, observability built in, human override at every level, auditable by design.

I am starting a PhD in AI/ML at George Washington University in Fall 2026, focused on advancing the science of agentic AI systems. I would welcome the opportunity to bring this combination of production agent engineering and defense-grade rigor to Anthropic's Agentic Systems team.

Sincerely,

Jhon B. Arango