Contract. Senior Software Engineer, Data Science (AI Accuracy) (NZ)

Work Environment

Location: Based in Auckland, New Zealand, with strong preference for candidates who can regularly work from our Auckland office alongside the Inspection and Progress AI teams.

Work model: Work In-Office - we support a hybrid approach and expect in‑office presence at least two days per week to stay closely aligned with engineers, PMs, and field-focused partners.

Collaboration: While your schedule offers flexibility, you'll maintain regular overlap with standard Auckland business hours (9:00 a.m.–5:00 p.m.) to enable effective collaboration with local and global teammates and support occasional on-call responsibilities.

Business Travel: Occasional domestic and international travel for team onsites may be required.

AI tooling: AI is built into how engineering work gets done at DroneDeploy. You’ll use AI-assisted tools for coding, data exploration, prompt iteration, and log/debug analysis, while remaining fully accountable for the quality, safety, and maintainability of your solutions.

Responsibilities

Build and maintain robust dataset and evaluation infrastructure, including ground-truth quality controls

Diagnose and fix measurement-path issues between offline evals and production accuracy

Iterate on and scale existing labeling frameworks and knowledge capture systems

Own prompt optimization across its full lifecycle, from candidate selection through production rollout and post-launch validation

Automate eval pipelines and build the tooling to run this at scale across multiple AI products

Create reusable internal tooling for error analysis, audits, and experimentation, and prioritize work using likely impact

Requirements

Strong professional experience with Python for data manipulation, analysis, and tooling (e.g., pandas, NumPy)

Solid SQL and database experience: querying, data modeling, and working with large datasets

A data science background with a knowledge of LLM accuracy and in particular visual reasoning of these models

Building and maintaining evaluation datasets, running and debugging evals with tools such as Braintrust or Langfuse, and conducting error analysis

Experience designing ground-truth and labeling systems, and measuring label agreement and data quality

Familiarity with prompt engineering and the full prompt lifecycle, from candidate selection through deployment and post-launch validation

Ability to reason about ontology and schema design and its downstream impact on models, evals, and tooling

Domain knowledge in construction or industrial inspection is a plus

A collaborative mindset and a preference for iterative, team-oriented development

Comfortable using AI-assisted tools for coding, data exploration, and debugging, while applying strong engineering judgment to guard against hallucinations and maintain high code quality.

Drone Certification: Not required for this role.