Senior Software Engineer, Huntress

• 6+ years of professional software development experience, including building and operating production backend systems
• Experience building applications with LLMs, including tool use, structured output, prompt and context design, and evaluation. Experience shipping agents to production is a strong plus
• An understanding of how LLM-based systems fail, and good habits for testing and debugging non-deterministic behavior
• Solid backend fundamentals, including APIs, background jobs and queues, concurrency, relational databases such as Postgres, and data stores such as Redis
• Strong skills in one or more backend languages and the ability to pick up new ones quickly. Our primary stack is Ruby on Rails. Prior Rails experience isn’t required, and we expect you to become fluent in it
• Security awareness: you think about untrusted input, access control, and keeping one customer’s data separate from another’s
• Curiosity about how security analysts think and work, and a willingness to spend real time learning from them
• Experience with AWS, Azure, or other public cloud environments
• Experience with AI coding tools, such as Claude Code
• BS or MS in Computer Science or Engineering, or equivalent experience

Desirable

• Reports to: Director, Engineering
• Huntress is building an Agentic SOC. When our detection pipeline raises a signal, an LLM-based agent picks it up and investigates it end-to-end. It gathers telemetry across our products, reconstructs what happened, and makes a well-documented determination about whether the activity is malicious
• Over time, it will take on more of the work itself, growing from resolving routine cases to acting fast when it finds an attack and handing analysts cases with the evidence already assembled
• This system is meant to augment our analysts, not replace them. It takes the high-volume, repetitive work off their plates so they can focus on complex cases and hands-on keyboard intrusions where human expertise matters most. It also reduces toil in the rest of their day by using LLMs to pre-process the output of the tools analysts use constantly, so they start from a clear picture instead of raw telemetry
• This isn’t a chatbot bolted onto a dashboard. The system investigates a large volume of signals a day across thousands of customer organizations. Many signals are benign, but some are live intrusions where minutes matter
• The telemetry the system reads comes from environments where attackers are actively operating. Because the system makes real decisions, like closing an investigation or isolating a host, it has to be right, it has to show its work, and it has to know when to hand off to a person
• As a Senior Software Engineer on the Agentic SOC team, you’ll build and ship core parts of this system. You’ll work closely with the team’s technical leads, SOC analysts, product researchers, and product managers. You’ll own significant pieces end-to-end, from an ambiguous problem to a production feature you can show is working
• We make full use of AI coding tools to enable you to accelerate your development. We don’t cap token usage. We expect you to use AI and all tools responsibly
• Build the investigation system. You design, build, and ship the components that make agentic investigation work: the tools agents use to query telemetry, the logic that drives an investigation, and the structured, evidence-backed output that analysts and customers see
• Turn analyst expertise into agent behavior. You sit with SOC analysts and product researchers to understand how they triage and investigate a given kind of Signal, then encode that into what the agent looks for, what it gathers, and how it explains its conclusions
• Take toil off analysts’ plates. You build LLM-powered features that pre-process the output of tools analysts use every day. They summarize, correlate, and highlight what matters, so analysts spend their time on judgment instead of assembly
• Measure what you ship. You build and extend the evaluations that tell us whether the system is getting better or worse, and you back your changes with evidence rather than a handful of good-looking examples
• Build with a security mindset. The agent reads attacker-influenced data, runs across thousands of customer tenants, and can take action in customer environments. You think about how what you build could be misused or manipulated, and you design for it
• Build for production. You write reliable, observable, cost-conscious code that behaves well when models are slow, wrong, or unavailable. You operate what you build and help keep the system healthy as new detections and products come online
• Collaborate and raise the bar. You give thoughtful code reviews, help teammates get unstuck, share what you learn about building with LLMs, and contribute to technical discussions and design reviews with clear reasoning about tradeoffs
• Balance rigor with pragmatism. You know when to prototype quickly to learn something and when to slow down because the cost of being wrong is high