November 2022 Summaries

Observability is crucial in both DevOps and machine learning contexts, where it involves monitoring workloads by accessing logs, application traces, and resource usage metrics. In machine learning, observability is vital for inspecting and debugging training pipelines, which interact with numerous third-party services and can incur substantial costs. Sematic, an open-source Continuous Machine Learning platform, enhances pipeline observability by surfacing logs, exceptions, and failures directly in the user interface, and by integrating with Grafana for resource usage monitoring. This allows for efficient detection and resolution of inefficiencies, such as idle GPU usage due to I/O bottlenecks, and helps manage overall costs by providing granular visibility into cloud expenditures. Additionally, Sematic supports the monitoring of inference servers to ensure they meet performance expectations and provides insights into model performance over time, potentially indicating when retraining is necessary. By abstracting infrastructure concerns, Sematic enables machine learning teams to quickly resolve issues and iterate on their models, facilitating a smoother development cycle.

Lineage Tracking is an essential feature for managing machine learning (ML) pipelines, ensuring that all assets consumed and produced during ML tasks are systematically tracked, thereby maintaining a reliable source of truth. This process is crucial for bookkeeping, debugging, traceability, compliance, and reproducibility, as it allows ML engineers to trace back all input parameters and configurations that lead to a particular model's outcome. Instead of manually logging data in spreadsheets or notebooks, which is cumbersome and inefficient, Lineage Tracking should be integrated into the ML platform itself. Sematic, an open-source Continuous Machine Learning platform, simplifies this by allowing users to define end-to-end ML pipelines in Python without requiring infrastructure skills, while inherently providing comprehensive traceability and lineage tracking. Sematic tracks various components such as code, configurations, input data, and resources, ensuring that the reproducibility of ML pipelines is easily achievable.

Sematic has announced a $3 million seed funding round led by Race Capital, with participation from Y Combinator, Soma Capital, Leonis Capital, Fundament, and Pioneer Fund, to advance its mission of making machine learning (ML) development more accessible and efficient. The company aims to revolutionize the ML industry by offering an open-source framework that simplifies the prototyping, automation, and deployment of end-to-end ML pipelines, addressing the current gap between theoretical potential and practical application due to complex tooling. Drawing from their experience at Cruise, the first robotaxi company with a commercial service, Sematic's founding team has developed methods that empower ML engineers to independently manage complex training pipelines. The company's vision includes creating automated feedback loops with guarantees around traceability, reproducibility, and observability, ensuring models remain relevant and scalable. Sematic seeks to become the standard for Continuous Machine Learning, providing a low barrier to entry and supporting ML engineers of all skill levels in building scalable, continuous learning systems.