Why Auctor
Auctor is building the AI layer for professional services and software implementation. Think of us as the brain behind the best solution engineers, forward-deployed engineers, and onboarding teams—automating the documentation, the discovery, and the decision-making that powers $400B+ in services work. We're going after one of the biggest software categories of the decade.
The Opportunity
As a Software Engineer, Applied AI at Auctor, you will design, build, and improve the core systems behind our agents in production.
This role sits at the boundary of engineering and empirical research. You will work across retrieval, document understanding, tool use, context management, prompting, and orchestration. Some weeks you will be shipping new capabilities. Some weeks you will be mining production traces, designing evals, and figuring out which part of the system is actually failing.
We are not looking for someone to glue an API onto a product and call it AI. We are looking for someone who wants to build real agent systems, understand how they behave in the wild, and use that understanding to make bold product and architecture decisions.
This role is based in New York, NY, in person 5 days per week.
What You'll Build
Build and improve the core systems behind our agents across retrieval, tool use, document understanding, memory, and orchestration
Design evals and experiments that help us understand agent quality in production
Turn traces, failures, and user behavior into concrete product and architecture decisions
Work closely with operations, GTM, and deployed teams to understand real workflows and where agents break down
Evaluate models, prompts, and system designs across real enterprise tasks
Own the loop from idea -> implementation -> measurement -> iteration
What You Bring
Strong engineering fundamentals and the ability to ship production systems
Fluency in Python
Experience building or working on LLM-powered products, agent systems, or adjacent applied AI systems
An empirical mindset — you reach for logs, traces, experiments, and real usage before guessing
Strong systems taste — you understand that retrieval, prompting, memory, tools, and UX interact
High ownership and comfort working in ambiguity
Strong opinions about what makes agent systems actually work