Senior Autonomous Security Engineer - Remote/Hybrid

AI Chopping Block; Northern (KY); Hybrid; USD 169,000 - 208,000; Full time

Benefits offered by this job

Hybrid & Remote Work

Competitive Compensation & Equity

Inclusive Team Culture

Job summary

Horizon3 is a fast-growing, remote cybersecurity company building an autonomous, black-box web application penetration tester. The role focuses on turning offensive expertise into an autonomous agent, with deterministic behavior and reliable proofs of findings.

You will own the attack-agent layer, implement orchestration, tool use, and evaluation to ensure scalable, unattended pentesting in production on AWS Bedrock, with strict production-safety and no false positives.

Qualifications

• 5+ years building production software, with strong Python.
• Hands-on experience building LLM-powered applications or agents with tool use and structured outputs.
• Track record of making LLMs reliable in production and shipping where it mattered.
• Experience with evaluation harnesses and measuring model/agent improvements.
• Strong instincts for prompt and context engineering and keeping models scoped.

Responsibilities

• Build and evolve the agent harness and orchestration.
• Design tools and feedback for the agent to probe and exploit with validation layers.
• Translate offensive expertise into repeatable agent capabilities.
• Own and grow the evaluation infrastructure and failure-mode taxonomy.
• Manage LLM inference in production: model selection, prompts, and cost control (AWS Bedrock).
• Ensure production-safety and no-false-positives in all findings.

Skills

Python

LLM-powered apps

Structured outputs

Tool integration

Production reliability

Prompt engineering

Observability

Ownership

Tools

AWS Bedrock

Neo4j

Job description

Horizon3 is a fast-growing, remote cybersecurity company building an autonomous, black-box web application penetration tester. The role focuses on turning offensive expertise into an autonomous agent, with deterministic behavior and reliable proofs of findings.

You will own the attack-agent layer, implement orchestration, tool use, and evaluation to ensure scalable, unattended pentesting in production on AWS Bedrock, with strict production-safety and no false positives.