Fieldguide
Software Engineer, (Internal Audit) [All Levels]
About this role
Fieldguide is hiring full-stack software engineers to build autonomous agents for SOX audit and internal control testing. You'll own end-to-end features that automate auditor workflows, ensure agent reliability and trustworthiness, and collaborate with domain experts to ship production AI systems that solve real audit problems.
What you'll do
- Design and build agent-driven experiences for SOX control testing, balancing automation with auditor oversight
- Own complex end-to-end workstreams across full stack, from agent logic to audit deliverables
- Improve reliability, repeatability, performance, and observability of long agentic runs
- Transform agent outputs into structured audit workpapers that adapt to firm-specific templates
- Collaborate daily with product, design, and SOX subject matter experts on domain-driven problems
- Help shape technical direction for agentic systems and engineering practices
What they're looking for
- Full-stack software development and shipping production code
- Building with LLMs in production (not just prototypes)
- Agentic systems and AI orchestration
- Comfort with ambiguity and 0→1 product development
- High ownership and self-direction
- Cross-functional collaboration and communication
- Ability to learn complex domain concepts quickly
- Familiarity with AI coding tools
Opens the application — the Jobs AI extension fills it for you. Set up autofill
Opens the official application on the employer’s site. No login required.
Fieldguide
Fieldguide builds audit and advisory software that modernizes trust and compliance for global commerce, with a focus on cybersecurity and privacy. The company is hiring Software Engineers, AI Engineers, and Production Support Engineers across all levels to design impactful features, develop agentic AI workflows, and drive system reliability.
- Website
- fieldguide.com
Likely interview questions
- Describe a production LLM system you built—what broke when real users and real data hit it, and how did you fix it?
- How would you approach making an agentic run produce the same defensible audit conclusion repeatedly, and what observability would you build in?