Christoph Stumpf

https://christoph-stumpf.com

Built a scalable, event-driven AI inference and search platform based on Kubernetes: • The system processes large volumes of image and video assets, enabling real-time search and retrieval of assets based on visual content. • The system is highly scalable and fault-tolerant, supporting distributed processing of assets and parallel inference via Nvidia Triton. • Asset searches are distributed using Milvus, a high-performance vector database, to achieve low-latency search results across millions of assets.

• Played a key role in the design and implementation of NOMAD AI, an innovative no-metadata, AI-based search engine used by The Associated Press (AP). • Worked on asset processing, real-time and batch inference and low-latency asset search. • Implemented a semantic video search feature to search for video scenes. • Built a face recognition system for video content. • Trained and fine-tuned large-scale ML models. • Designed a highly scalable ML inference pipeline on AWS, processing millions of assets.

• Technical and organizational lead of the machine learning team. • Management of the machine learning platform, the ML operations process and the data management system.

• Designed an E2E-ML platform for model training, deployment and monitoring. • Implemented deep learning based computer vision models to classify content in web pages. • Built a machine learning model to detect patterns of interest in JavaScript source code. • Productionized machine learning models using MLflow, BentoML, Flask and Docker. • Designed data ingestion and data processing pipelines using Spark and Delta Lake. • Managed operations of cloud platforms including Azure, AWS and Google Cloud Platform.

• Studied machine learning fundamentals and theoretical background. • Received certifications in MOOCs focused on machine learning and deep learning. • Completed the Springboard Machine Learning Engineering Career Track • Capstone Project - Machine learning based Intrusion Detection System: Implemented a classifier capable of categorizing network flows as either benign or malicious in order to detect network attacks and intrusions.

• Designed and implemented the API for an object detection service. • Built a distributed messaging infrastructure for event-driven application communication of 10+ services. • Led the development of an reactive microservice framework used as foundation for all client-facing services company-wide. • Led the design and development effort for multiple services within a large microservice ecosystem. • Built a high-speed data-upload service ingesting millions of video, audio and image files.

• Led multiple projects as the technical lead. • Guided the application development team in selecting development tools and programming frameworks. • Led the development team implementing a RESTful web API for fund-related data. • Built a reporting application for private banking customers integrated in the online-banking solution.

Built graphical and markup-based tools on the .NET platform to increase in-house developer velocity.

Developed the primary fund management platform of Erste Asset Management for in-house clients.