Experience | Dhia

Designed and shipped a Go microservice on top of the core observability stack (Prometheus, Thanos, Grafana) that profiles each customer’s metric behavior, detects genuine anomalies, and automatically recalibrates alert thresholds through a continuous feedback loop.
Eliminated the false positives and missed incidents caused by static, hand-tuned thresholds, cutting alert noise by 40% across all customer environments.
Built a metric profiling and calibration tool that auto-generates deployment artifacts from historical data, and delivered the full solution to production Kubernetes clusters through a GitOps pipeline.

Built an in-house ETA prediction model that estimates driver travel times between pickup and drop-off points, cutting average prediction error by 45 seconds against Google Maps.
Automated the full retraining lifecycle using Apache Airflow, keeping the model continuously updated as new trip data comes in.
Delivered an interactive Power BI dashboard giving operations managers a real-time view of driver and company performance.

Developed a real-time weapon detection system for video surveillance using deep learning models (YOLOv8, Faster R-CNN, Mask R-CNN) and OpenCV.
Benchmarked and compared multiple detection architectures to achieve high accuracy and low latency for security applications.
Implemented an end-to-end data pipeline, from image preparation and labeling to model deployment and real-time inference monitoring.
Deployed and maintained models, evaluating performance and ensuring system stability using Jupyter and Git.