Agent Harness Engineer - Core Runtime

About NinjaTech AI

NinjaTech AI is an innovative generative AI startup with headquarters in Silicon Valley and offices in Sydney and Vancouver. Supported by Alexa Fund and Samsung Ventures, the company develops autonomous AI agents designed to streamline real-world tasks using generative AI technology. Their leading product, SuperNinja, offers a comprehensive AI platform featuring capabilities like website creation, full-cycle coding, advanced data analysis, and more.

Role Overview

We are seeking a Core Runtime Engineer responsible for developing the agent harness that transforms large language models (LLMs) into dependable, autonomous software entities. This harness comprises the run loop, orchestrator, and monitoring components critical to maintaining continuous, safe operation of agents across thousands of tool invocations over extended durations. The role bridges systems engineering and applied AI, forming the foundational substrate for every SuperNinja agent.

Key Responsibilities

• Design and continuously improve the agent runtime loop, focusing on context management, turn compression, and recovering from incomplete or failed tool executions.
• Develop an orchestrator capable of launching, supervising, and managing long-running background agents while preserving their state.
• Implement a monitoring system that identifies issues like lockups, infinite loops, or runaway agents and intervenes without disruption.
• Create and maintain a provider abstraction layer that accommodates various model backends such as Claude, Codex, ensuring the harness remains model-agnostic.
• Build end-to-end observability with structured telemetry for efficient debugging and fault tracing in production environments.
• Optimize for performance and cost-efficiency at scale to support millions of daily user interactions.

Qualifications

• A Master's or Bachelor's degree in Computer Science or equivalent hands-on experience.
• At least 4 years experience developing production backend or systems software, primarily using Python.
• Solid understanding of concurrent programming, process supervision, and fault-tolerant system design.
• Practical exposure to large language model APIs and agent frameworks involving tool and function calling along with streaming.
• Demonstrated history of successfully deploying and operating highly reliable services at scale.

Role Details

• Report to the engineering lead responsible for agent infrastructure.
• Lead research, design, and construction of the core run loop and orchestration mechanisms powering autonomous agents.
• Determine methods for managing agent context, error recovery, and smooth task handoffs.
• Create prototypes and proof of concepts to rapidly validate and evolve harness features into deliverable products.
• Keep abreast of current advances in agent architectures and incorporate best practices into the runtime.
• Develop self-healing capabilities to enable graceful degradation rather than abrupt failure.
• Collaborate cross-functionally with teams in applied science, product management, and platform integrations.

Desirable Experience

• Previous experience building agent frameworks, workflow engines, or systems for managing long-running jobs.
• A strong focus on reliability akin to site reliability engineering (SRE) principles, including comprehensive instrumentation.
• The ability to handle unpredictability in LLM outputs and architect systems around such non-determinism.
• Familiarity with sandboxing, container orchestration, or approaches to secure code execution.
• Ability to quickly develop prototypes and determine when to mature them for production scalability.

Why Join NinjaTech AI

This opportunity allows you to develop the foundational harness on which all autonomous agents operate. The success of AI autonomy depends heavily on this runtime layer, and you will have full ownership working closely with a team that balances strong engineering discipline with scientific exploration.