Sebastian Keil

Optimized model latency for different hardware targets, such as AWS Inferentia and CPU, which enabled the integration of large-scale Machine Learning on computationally contrained, client-specific environments despite fluctuating, large-scale data volumes. Owned the end-to-end integration state-of-the art NLP methods, including private-hosted LLMs, into the core software product; engineered the backend architecture for AI capabilities while maintaining strict security protocols. Orchestrated containerization of the infrastructure, ensuring environmental parity across development and production. Developed automated Jenkins pipelines that transitioned the deployment process from days ofmanual labor to a single-click operation.

Developed a BERT-based classification system to automate company categorization via web-scraping; implemented distributed training on AWS SageMaker, significantly accelerating database expansion and lead qualification.

Developed dimensional sentiment analysis techniques to model excess returns in financial markets; developed Active Learning strategies for efficient labeling, significantly reducing the manual annotation effort required for domain-specific financial datasets.