PodcastSmall Data, Big Impact: The Story Behind DuckDBThreads / The platform team / 2024
The platform team in 2024
29 sessions
PodcastLightweight Feature PlatformClaimTecton is a feature platform that decouples feature engineering code from models and helps manage feature pipelines.2:57
PodcastDesigning ML Infra for ML & LLM Use CasesClaimA platform's North Star should be enabling data scientists to move an idea from ideation to production as quickly as possible.15:35
PodcastWhat Business Stakeholders Want to See from the ML TeamsClaimPeter Guagenti built the Acquia Lift personalization and targeting product despite internal resistance, and the product later became a major part of the company’s business.27:14
AI Innovations: The Power of Feature PlatformsPushed backNikhil Garg says a feature store and a feature platform are different: a feature store mainly provides storage and serving, while a feature platform also includes computation and other end-to-end capabilities.44:51
Navigating the Emerging LLMOps StackClaimHien Luu leads the ML platform team at DoorDash and is sharing learnings from exploring the LLMOps stack while considering a strategy for building an internal stack.0:30
PodcastUber's Michelangelo: Strategic AI Overhaul and ImpactClaimUber’s Michelangelo platform went through three phases: foundational predictive machine learning from 2016 to 2019, deep learning from 2019 to 2023, and generative AI from 2023 onward.4:56
Data Quality = Quality AIClaimPushkar Garg recommends implementing data-quality checks at the platform level and making operators and tools available in data pipelines for data scientists.20:27
PodcastDesign and Development Principles for LLMOpsClaimAndy McMahon says AI and traditional machine learning should be treated as capabilities in one ecosystem rather than as completely separate platforms.31:12
PodcastWho's MLOps for Anyway?ClaimPeople, process, and platform all matter in MLOps, but a platform cannot compensate for unclear processes.24:41
PodcastGlobal Feature Store: Optimizing Locally and Scaling Globally at Delivery HeroPushed backThe participants disputed whether one centralized machine learning platform should define tooling and priorities for all departments.4:03
PodcastReinvent Yourself and Be CuriousPushed backStefano Bosisio disputes the assumption that technically impressive ML platforms will be adopted automatically, arguing that internal communication and education are necessary.18:48
PodcastCentralized or Decentralized ML Platform?ClaimPicnic currently has a platform team and two broad machine learning domain teams, one focused on consumer use cases and one focused on supply-chain and operations use cases.46:49
Chronon: Airbnb's Open-Source Data PlatformClaimChronon is a data platform for machine learning that was built at Airbnb and Stripe.0:32
Engineering Your AI PlatformClaimAI platforms must handle unstructured data, which is larger and more complex than structured tabular data.3:15
How Feature Stores WorkClaimA feature platform or physical feature store handles transformation and related problems, but products in this category often impose proprietary compute engines or domain-specific languages.16:55
The Evolution of Lyft's Feature StoreClaimLyft's feature service is used by machine-learning models, the marketing and communications platform, driver incentives, fraud detection, and dispatch.2:08
The Future of Data: Composability & the Modular Data StackClaimWhatnot's data platforms team is scaling an architecture that supports AI and machine learning, applications, and analytics use cases.1:13
How to Create a Multi-Agent AI System in JavaScriptClaimJavaScript is widely adopted by enterprises and runs in browsers, servers, and other devices.2:39
Building an ML Platform from scratchPushed backBen says Prefect is the one tool he believes is nearly always useful, while other platform tools can be overkill for a single person; Eric's discussion emphasizes starting with simpler monolithic pipelines.1:37:43