Pittsburgh TechFest 2026

Most engineering teams approach AI opportunistically, treating it merely as a faster way to write code. However, software development represents only 20% of the software value stream, and localized optimization frequently introduces systemic chaos downstream during testing, review, and deployment.

This hands-on session builds directly upon real-world production methodology to demonstrate how semi-autonomous, role-based AI agents can be structured across the entire lifecycle—from product definition and architectural design decisions (ADRs) to automated builds and containerized cloud operations. Using a concrete energy grid intelligence case study (GridPulse), we will dismantle the typical "vibe-coding" approach and replace it with an engineering discipline focused on structured context architecture, explicit API contracts, and spec-driven development

Connecting machines to the cloud is only the first step toward intelligent manufacturing. This presentation shares the technical journey of redesigning Conair's SmartServices platform to support large-scale industrial monitoring, remote machine control, and future AI-driven insights.
We'll cover the end-to-end architecture—from industrial protocols and Azure IoT Edge through cloud services and real-time dashboards—and discuss the tradeoffs between edge and cloud processing, telemetry retention, and operating costs. Along the way, we'll examine how a unified equipment model enables diverse machine types to be monitored through a common interface.
The session concludes by exploring how the platform enables predictive maintenance, anomaly detection, and optimization using AI, along with the practical challenges of training and deploying machine learning models in industrial environments.