Machine Learning Engineer

Worked at a research and engineering institute at NTUA delivering applied AI and computer vision solutions for EU-funded security and civil-protection programmes. Independently led computer vision project delivery from problem definition through deployed systems.

Built real-time UAV computer vision pipelines for object detection, multi-object tracking, and re-identification on NVIDIA Jetson edge devices using YOLO, RT-DETR, MeMOTR, ByteTrack, OpenCV, GStreamer, ONNX, Docker, CUDA, and MQTT. Implemented tracked-object geolocation from UAV telemetry, velocity estimation, geofencing, production Postgres integration, and end-to-end ML workflows including dataset sourcing, ground-truth evaluation, and fine-tuning.

Developed a hybrid content recommendation engine using session-based association mining and embedding similarity, served through FastAPI with scheduled retraining. Added Prometheus inference metrics and Grafana dashboards, and built a body-pose estimation pipeline for emotion recognition while prototyping synthetic data generation in NVIDIA Omniverse.

Developed transfer-learning neural networks to simulate multiple engine types, reducing development time for new engine variants and improving deployment speed and model performance.

Applied pattern-detection and change-point-detection methods to operational time-series data to improve engine performance analytics.

Built and maintained ETL pipelines and backend integrations for ActionIQ, an enterprise customer data platform. Supported ingestion and export workflows for international clients.

Developed processing and orchestration capabilities with Apache Spark and Luigi, extending an existing pipeline task system. Wrote and modified Terraform infrastructure configurations, supported CI/CD through GitHub Actions and Datadog monitoring, and contributed to a greenfield .NET backend service for four months.

Built and shipped a neural-network solution to extract and classify email signature blocks from real-world customer data.

Containerized the service with Docker and deployed it to Azure as a running production service.

Completed a five-year integrated programme worth 300 ECTS, equivalent to Bachelor's and Master's level study.

Completed a diploma thesis titled Application of Reinforcement Learning Algorithms in Sonic the Hedgehog™, implementing multiple reinforcement learning algorithms from scratch and achieving results comparable to OpenAI baselines.