Published inData Engineer ThingsRebuilding Redis on S3 ExpressA slow, cheap and scalable key-value storeFeb 2, 2024A response icon2Feb 2, 2024A response icon2
Published inCantor’s ParadiseMarginally Important: Automatic Differentiation with Dual NumbersA brief (and inaccurate) history of derivatives, with a brief (and incomplete) Python implementationJan 9, 2024A response icon1Jan 9, 2024A response icon1
Published inThe TechlifeApplied Research at Reasonable ScaleBuilding ML-heavy products without breaking the bankNov 15, 2022A response icon1Nov 15, 2022A response icon1
Published inTDS Archive“Is This You?” Entity Matching in the Modern Data Stack with Large Language modelsAn experiment in productionizing LLMsOct 27, 2022A response icon1Oct 27, 2022A response icon1
Published inTDS ArchiveNDCG Is Not All You NeedBehavioral testing for recSys with RecListJan 12, 2022Jan 12, 2022
Published inCoveo Behind the ClickThe Vectors That Came in From The ColdImproving “Cold Start” at Scale in a Multi-tenant SaaS ScenarioDec 21, 2021Dec 21, 2021
Published inTDS ArchiveThe Post-Modern StackJoining the modern data stack and the modern ML stackSep 22, 2021Sep 22, 2021
Published inTDS ArchiveMLOps without much OpsOr: How to build AI companies leveraging open-source and SaaS softwareSep 22, 2021Sep 22, 2021
Published inTDS Archive“DAG Card” is the new “Model Card”Automated generation of DAG Cards from Metaflow, inspired by Google Cards for machine learning models.Mar 13, 2021Mar 13, 2021
Published inTDS ArchiveNoOps Machine LearningA PaaS End-to-End ML Setup with Metaflow, Serverless and SageMaker.Feb 10, 2021A response icon4Feb 10, 2021A response icon4