Senior Full-Stack Engineer, Autonomous Defense

Horizon3

8d•$195,000 - $235,000•Remote

About The Position

As a Senior Full-Stack Engineer on Horizon3’s Autonomous Defensive team, you will build the platform that NodeZero’s defensive agents run on. You will turn AI researchers’ agent capabilities into reliable production systems that operate across customer tenants and environments with strict isolation, credential, and egress constraints. The platform must safely simulate proposed production changes, stage them, require approval where appropriate, support rollback, and maintain an auditable record of what happened and whose authority was used. This is an end-to-end platform role spanning the agent runtime, tool layer, orchestration, connector framework, observability, and operator-facing UI. You will build backend systems and React and TypeScript interfaces that help users understand agent behavior and intervene confidently. You will partner with AI researchers, attack engineers, and product teams to move capabilities from prototype to production, while communicating platform constraints early. The role focuses on building the systems that make agent work shippable and safe, rather than tuning prompts.

Requirements

Bachelor’s degree in Computer Science, Computer Engineering, or a related field.
At least 7 years of professional software engineering experience using modern object-oriented or functional languages such as Python, Go, Scala, C++, or TypeScript.
Proven experience delivering backend APIs, especially GraphQL, and consuming them from a modern web stack.
Production experience with React, TypeScript, and modern CSS layout, including the ability to build usable interfaces for complex workflows.
Experience building applications on cloud platforms such as AWS, Azure, or GCP, using container technologies such as Docker and Kubernetes.
Expert proficiency in SQL.
Demonstrated experience designing and operating distributed systems in production, including asynchronous workflows, queues, retries, idempotency, and partial failure.
Track record of shipping and operating production services, including on-call ownership.

Nice To Haves

Experience building agentic or LLM-backed systems in production.
Experience with a durable execution or workflow engine such as Temporal, or having built equivalent checkpointing.
Experience building multi-tenant platforms with hard isolation requirements, workload identity, and short-lived credential brokering.
Experience integrating numerous third-party APIs behind a normalized abstraction.
Familiarity with security control planes and their APIs, including EDR products such as CrowdStrike, SentinelOne, and Microsoft Defender, as well as firewalls, SIEM, and cloud IAM.
Experience with relational and graph database architectures, including PostgreSQL and Neo4j, and building GraphQL backends.
Experience with client-side data and streaming patterns such as React Query or Apollo, GraphQL subscriptions, and WebSockets for long-running, incrementally updating state.
Experience building operator-facing or admin interfaces where density, auditability, and safe handling of destructive actions matter more than visual polish.
Experience with observability tooling such as Datadog, Prometheus, or Grafana, and distributed tracing.
Awareness of prompt injection and untrusted-input handling in systems where an agent consumes data from its operating environment.

Responsibilities

Build and own the agent execution runtime, including durable, resumable long-running workflows with checkpointing, idempotent steps, cancellation, and step- and run-level timeouts.
Design and implement an agent tool registry with versioning, schema validation, per-tenant enablement, and server-side authorization.
Build a connector framework for third-party security control planes, including EDR, firewall, identity, SIEM, and cloud IAM systems.
Implement a safe-change pipeline with dry runs and simulation, blast-radius classification, approval gates, staged rollout, rollback, and immutable per-tenant audit trails.
Build operator-facing interfaces in React and TypeScript that make complex system state understandable and actionable.
Build client-side data and streaming capabilities that keep interfaces accurate against long-running, incrementally updated state using WebSockets rather than polling.
Extend NodeZero’s tenant isolation model to agents taking write actions, addressing execution boundaries, egress control, quotas, and cross-tenant leakage in caches, indexes, and logs.
Develop product features in ETL and GraphQL to support data processing and retrieval for agent context, run history, and remediation state.
Build ETL pipelines that turn pentest and remediation outcomes into datasets for researcher training and evaluation.
Instrument systems with distributed tracing across agent runs, per-tenant cost and token accounting, and tools for explaining agent behavior.
Partner with AI researchers, attack engineers, and product teams to take capabilities from prototype to production and provide early feedback on platform constraints.