Selected works demonstrating our capability to deploy complex AI architectures in real-world industrial and clinical environments.
**The Challenge:** A major energy conglomerate needed to track raw material inventory across a massive yard. Manual counting was slow and error-prone.
**The Architecture:** We deployed a drone-based feed analysis system using YOLOv8 for object detection and DeepSORT for tracking. The system segments materials in real-time, estimating volume and updating a central inventory database automatically.
**The Challenge:** Measuring systolic and diastolic blood pressure typically requires a cuff. We sought to predict these vitals using only simple PPG optical signals (smartwatch data).
**The Architecture:** We engineered a deep regression model trained on 8,000+ patient records. By reducing demographic bias and optimizing for ANSI standards, we achieved a Standard Deviation of < 9.5mmHg.
**The Challenge:** Detecting Peripheral Artery Disease (PAD) often happens too late. Early detection via ABI (Ankle-Brachial Index) is crucial but often overlooked.
**The Architecture:** Utilizing the HeartGPT framework and fine-tuned tabular transformers, we developed a risk scoring engine. This model analyzes ordinal bins of patient data to predict disease severity with high sensitivity.