AI Research Engineer

Twine builds AI Digital Employees who help cyber teams close the execution gap. Our first AI Digital Employee, Alex, learns, understands and takes away the burden of Identity and Access Management (IAM) tasks - proactively completing the organization's cyber objectives.

About the Role:

You'll own technically difficult problems before the solution is known. Starting with a cybersecurity need, you'll study the domain, frame the technical challenge, investigate possible approaches, and build the solution through production. You'll have access to Twine's shared agent infrastructure, but the role goes far beyond configuring existing components. Many use cases will require new reasoning methods, agent architectures, evaluation techniques, data strategies, or model behavior. This is a role for someone who enjoys applied research but is motivated by outcomes. The goal is not simply to prove that an idea can work, it is to make it useful and reliable for customers.

What You'll Do:

• Design and implement AI solutions for complex cybersecurity workflows.
• Develop new approaches for agent reasoning, planning, tool use, context management, decision-making, and recovery from failures.
• Determine the right technical approach for each problem, combining models, algorithms, data, and conventional software where appropriate.
• Define quality for each use case and create evaluations that measure correctness, task completion, consistency, safety, latency, and cost.
• Analyze agent behavior deeply, identify the underlying causes of failures, and improve the relevant model, data, tool, or architecture.
• Collaborate with cybersecurity experts, product teams, software engineers, and AI infrastructure engineers throughout development.

Requirements:

• Excellent software engineering skills and substantial experience with Python.
• Hands-on experience building advanced systems with modern language models.
• Deep familiarity with several areas relevant to AI agents, such as planning, tool use, retrieval, memory, structured generation, model adaptation, or evaluation.
• The ability to turn an ambiguous problem into a technical research plan and then into working software.
• Strong experimental instincts: you form clear hypotheses, design meaningful tests, and make decisions based on evidence.
• Sound judgment about when to use an AI technique and when a deterministic approach is more effective.
• Ownership of the complete result, including its behavior after reaching production.

Nice to Have:

• Developing autonomous or long-running agents that interact with real systems.
• Building evaluation environments, simulations, or benchmarks for complex model behavior.
• Experience in cybersecurity, identity, security operations, or enterprise automation.
• Applying published research or your own research to production problems.

‍

‍