AURRA is an AI-powered smart thermostat platform built as a documentation vehicle — a technically realistic IoT product designed specifically to demonstrate API reference documentation, developer integration writing, SDK documentation, and consumer-facing UX writing in a single cohesive suite.
Several engagements in this portfolio — particularly NYC OCME — produced technical documentation that cannot be publicly shared under NDA. AURRA was built to close that gap: a self-directed project with a technically realistic product, designed from scratch to exercise the full range of technical documentation disciplines.
The product scenario is intentionally demanding: an IoT thermostat platform with a REST API, OAuth 2.0 authentication, predictive HVAC diagnostics, webhook infrastructure, and multi-platform SDK integration. Every documentation challenge present in enterprise work — API taxonomy, developer onboarding, end-user cognitive load — is represented.
All four artifacts are original. No source material was adapted or transcribed. The structural decisions, information architecture, and content strategy choices reflect the same methodology applied across Domino's, Google, and DoorDash engagements.
Each document targets a distinct reader — API consumers, integration engineers, end users, and ML stakeholders — demonstrating the ability to calibrate register, structure, and information density across the full documentation spectrum.
Complete endpoint documentation for the AURRA REST API. Covers authentication, thermostat control, predictive diagnostics, and voice assistant integration — each endpoint with parameter schemas, example requests, response shapes, and Python + JavaScript code samples.
A developer-facing guide covering the full AURRA integration lifecycle — authentication setup, telemetry endpoints, predictive maintenance configuration, webhook registration, and SDK quickstart. Code-first, task-oriented, minimal friction.
Consumer-facing documentation for the AURRA platform. Covers device setup, scheduling, eco mode, predictive alerts, and app integration — applying plain language, progressive disclosure, and task sequencing calibrated to non-technical cognitive load.
Transparency documentation for the ML model powering AURRA's predictive HVAC maintenance. Covers model architecture, training data composition, SHAP feature importance, threshold calibration with cost-sensitive analysis, subgroup performance, limitations, and ethical considerations.