Geralt Rivia

Designed and deployed end-to-end ML pipelines for autonomous aerial systems, covering data collection, preprocessing, model training, evaluation, and inference in production environments. Co-developed a custom human–hybrid dataset creation pipeline, enabling scalable data ingestion, annotation automation, augmentation, and validation to support supervised learning workflows for computer vision models. Trained, fine-tuned, and optimized object detection and tracking models (YOLOv5/v8, custom CNNs), focusing on generalization, latency, and robustness under real-world conditions.

ML Engineer at Emvirt Solutions (2023-11 – Present)

• Designed and deployed end-to-end ML pipelines for autonomous aerial systems, covering data collection, preprocessing, model training, evaluation, and inference in production environments.
• Co-developed a custom human–hybrid dataset creation pipeline, enabling scalable data ingestion, annotation automation, augmentation, and validation to support supervised learning workflows for computer vision models.
• Trained, fine-tuned, and optimized object detection and tracking models (YOLOv5/v8, custom CNNs), focusing on generalization, latency, and robustness under real-world conditions.
• Built real-time ML inference pipelines using OpenCV, NVIDIA DeepStream, and custom plugins, exporting structured predictions (CSV/JSON) for downstream analytics and monitoring.
• Implemented model deployment and lifecycle workflows across edge–cloud systems, containerizing inference services with Docker and deploying on AWS (EC2, S3) with CI/CD support.
• Developed backend services in Golang and Python to manage model inference APIs, telemetry data, and dataset metadata, integrating MongoDB and WebSockets.
• Optimized ML workloads on edge hardware (Jetson Orin, Raspberry Pi 5 + Hailo), applying model compression, pipeline optimization, and hardware-aware inference strategies.

Bachelor of Technology in Computer Science – Rajasthan Technical University (2019-07 – 2023-07)