Senior AI Software Engineer
Summary
Senior hands-on engineer at Eversor AI in Auckland building and owning LLM-powered product features end to end — document extraction, retrieval, prompting and structured outputs — with accountability for accuracy, latency and cost in production. Core stack is Python plus a modern web stack; role also involves mentoring and leading workstreams.
Salary: $140,000 – $190,000 per year
About the role
This is a permanent, senior, hands on engineering role that will join us from 2027. You will build the AI features inside our products, from document extraction and retrieval through to the outputs a customer sees, and you will own them in production.
You will report to our technical lead and work alongside our product manager. Most of your week is building. The rest is what a growing team needs from its senior people: taking the lead on features and small workstreams, reviewing code, mentoring, and stepping up when the technical lead is away so the work keeps moving.
Key responsibilities
Build AI features from brief to production: document extraction, retrieval, prompting, structured outputs and the tests that prove they work
Own what you ship. Accuracy, latency and cost per call are part of done, and you keep them visible
Take the lead on features and small workstreams: plan them, split the work, review it, land it
Mentor the developers around you through code review, pairing and clear written decisions
Step up for the technical lead when they are away, so the AI work keeps moving
Work with our product manager and join customer conversations so what you build answers a real problem
Investigate production issues and fix the cause, not the symptom
Help with technical questions from customers and the commercial team when needed
About you
Several years building software in production, with the last few spent shipping LLM or vision model features that real users depend on
Hands on depth in at least one of: retrieval pipelines, structured extraction from documents, agentic workflows, evaluation of model outputs
Strong Python, plus working knowledge of a modern web stack
Evidence that other developers have got better because they worked with you
The judgement to choose between open weight and hosted models on accuracy, cost and data handling, and to explain the choice plainly
A habit of measuring before and after, and of saying so when something is not worth building
Clear written communication. Much of what you decide will be read by people who were not in the room
About us
Eversor AI builds products that read complex documents, check them against the rules that apply, and help organisations make decisions faster and with more confidence. Our customers work in industries that still run on paper, email and long checklists. We replace that with software they can trust. We are a small Auckland team with products in customers' hands, and we are growing.