engblogs

Creating testing environments using sports analytics to advance AI research beyond the lab

Describes how the SoundCloud API now exposes high-tier (Go+) content via a new access filter (playable, preview, blocked), enabling client-side UX control and updated track metadata annotations.

A concise guide to exposing high-tier (Go+) content via the SoundCloud API using an access filter to control playable, preview, and blocked states, with metadata annotations for playback accessibility.

Retrieving metadata for the full SoundCloud catalog by using the new filter in API requests.

Applying game theory to large-scale data analysis can help solve fundamental machine learning problems.

Exploring the implications of Will Hurd joining OpenAI’s board of directors

How SoundCloud's Corpus project turned scattered data into a governance-driven single source of truth on BigQuery by unifying ETLs into a usable, scalable data platform focused on quality, timeliness, and maintainability.

A case study of building Corpus, SoundCloud's data governance-driven journey to a single source of truth on BigQuery, underpinned by six guiding principles: quality, compliance, timeliness, usability, efficiency, and maintainability.

Exploring the process of using user interaction data to drive business decisions

Guides teams through planning, documenting, and executing successful internal system handovers using an RFC-driven guideline to ensure ownership, continuity, and reliable operations.

A practical RFC-based guideline for handing over system ownership during organizational change, detailing what to consider, how to plan, and how to document architecture, runbooks, and on-call processes to prevent outages and maintain continuity.

A technical exploration of Rotary Positional Embedding (RoPE), a novel encoding technique that combines absolute and relative approaches, including experimental results.