--- title: "Do we ask and answer all the Big Queries? Yes, we do!" newsletter: "MLOps Community" date: 2024-08-29 source: https://aaif.live/newsletters/mlopscommunity/2024-08-29-do-we-ask-and-answer-all-the-big-queries-yes-we-do --- # Do we ask and answer all the Big Queries? Yes, we do! *Plus: Is local always best? Do you have insights from this summer's must-read? How do you stay up to date with data? What's in a name? And more!* *MLOps Community — Agentic AI Foundation, 2024-08-29* https://learn.mlops.community/courses/languages/your-first-mlops-stack/ You'll be dancing too, when you see just how much has been packed into our new course, Your First MLOps Stack [https://learn.mlops.community/courses/languages/your-first-mlops-stack/]. Work with Apache Beam to build data pipelines, use Kubeflow to manage model training, apply MLflow for experiment tracking, and more. For those just starting out and those looking to refine skills, this course will help keep your stack on track. ## BigQuery Feature Store // Nicolas Mauti // MLOps Podcast #255 Should you decouple when using a feature store? That’s the BigQuery, big question. Nicolas and I discuss Malt's use of BigQuery as a feature store, emphasizing the benefits of decoupling feature engineering from model training and code development. This separation improved workflow efficiency, allowing data scientists to focus on one task at a time, reducing inconsistencies. We also explored challenges, like adding new features and real-time processing limitations, and anticipated future architectural changes to handle growing data and real-time feature computation. For all the big questions, you know this podcast gives you the big answers! Video [https://home.mlops.community/home/videos/bigquery-feature-store] || Spotify [https://open.spotify.com/episode/5bqG86R6XV9XKeIRiLIQpW?si=2258fd0b19fa4867] || Apple [https://podcasts.apple.com/us/podcast/bigquery-feature-store-nicolas-mauti-255/id1505372978?i=1000666392395] [https://podcasts.apple.com/us/podcast/bigquery-feature-store-nicolas-mauti-255/id1505372978?i=1000666392395](https://podcasts.apple.com/us/podcast/bigquery-feature-store-nicolas-mauti-255/id1505372978?i=1000666392395) [https://home.mlops.community/home/videos/bigquery-feature-store](https://home.mlops.community/home/videos/bigquery-feature-store) [https://open.spotify.com/episode/5bqG86R6XV9XKeIRiLIQpW?si=2258fd0b19fa4867](https://open.spotify.com/episode/5bqG86R6XV9XKeIRiLIQpW?si=2258fd0b19fa4867) ## Athina IDE ## 10x your AI development speed with Athina IDE https://athina.ai/?utm_source=sponsor&utm_medium=email&utm_campaign=mlops_community_collaborate Athina is a collaborative IDE that helps teams prototype, experiment, evaluate and monitor AI. Top AI teams like Perplexity, Meesho, and Leonardo use Athina IDE to collaborate on datasets, run experiments, and test prompts, models and retrievers 10x faster in a spreadsheet-like UI. * For engineers: Engineers can work with large datasets to engineer and test complex chains in a python SDK or in Athina’s UI. * For product managers: PMs can manage prompts, monitor LLM performance in production, and even run experiments without writing code. * For QA and Domain Experts: Human annotators can work alongside AI-assisted evaluation with a dedicated “annotation mode”. Athina IDE: [https://bit.ly/athina-ide-demo] Demo Video [https://bit.ly/athina-ide-demo] | Athina AI Website [https://athina.ai/?utm_source=sponsor&utm_medium=email&utm_campaign=mlops_community_collaborate] Blog: [https://hub.athina.ai/guides/building-an-ideal-tech-stack-for-llm-applications-from-scratch?utm_source=sponsor&utm_medium=email&utm_campaign=mlops_community_collab] Building an ideal tool-stack for LLM development → [https://hub.athina.ai/guides/building-an-ideal-tech-stack-for-llm-applications-from-scratch?utm_source=sponsor&utm_medium=email&utm_campaign=mlops_community_collab] ## MLOps for GenAI Applications // Harcharan Kabbay // MLOps Podcast #256 Craft fairs, boutique shops, local food produce – some things are best enjoyed locally, right? But what about LLMs? In this episode, we chatted about the challenges and best practices for operationalizing machine learning, especially with local LLMs and CI/CD pipelines. Harcharan talked about the importance of moving beyond local testing to a more structured, production-ready setup using Kubernetes, Argo CD, and other tools. We also discussed the risks of relying too much on local setups, the need for resilient architectures to avoid single points of failure, and the crucial role of observability in monitoring performance and security. After clicking "listen," head out to find a craft brewer – you know, to support local. Video [https://home.mlops.community/home/videos/mlops-for-genai-applications] || Spotify [https://open.spotify.com/episode/1PXKxzBIbNDj3ZpDWGKMLU?si=88d8c02a776f4ba7] || Apple [https://podcasts.apple.com/us/podcast/mlops-for-genai-applications-harcharan-kabbay-256/id1505372978?i=1000666813375] [https://podcasts.apple.com/us/podcast/mlops-for-genai-applications-harcharan-kabbay-256/id1505372978?i=1000666813375](https://podcasts.apple.com/us/podcast/mlops-for-genai-applications-harcharan-kabbay-256/id1505372978?i=1000666813375) [https://home.mlops.community/home/videos/mlops-for-genai-applications](https://home.mlops.community/home/videos/mlops-for-genai-applications) [https://open.spotify.com/episode/1PXKxzBIbNDj3ZpDWGKMLU?si=88d8c02a776f4ba7](https://open.spotify.com/episode/1PXKxzBIbNDj3ZpDWGKMLU?si=88d8c02a776f4ba7) ## Job of the Week [https://joinhandshake.com/careers/open-roles/job/?gh_jid=6210064](https://joinhandshake.com/careers/open-roles/job/?gh_jid=6210064) ## Exploring Long Context Language Models // MLOps Reading Group August 2024 During summer, there’s nothing quite like chilling on the beach with a good book—or even better, a research paper and a great discussion! This group explored the paper Can Long-context Language Models Subsume Retrieval, RAG, SQL, and More?, [https://arxiv.org/abs/2406.13121] which argues they can potentially outperform traditional methods by handling extensive text contexts more effectively. There was some acknowledgement in the group about the potential of LCLMs, but an agreement that current retrieval methods still hold value, especially considering the paper's focus on the Gemini model. Concerns about real-world applicability and the need for further validation were also raised in this great analysis. Look out for our next session coming soon, and if you have suggestions for future papers, reach out to us at binoy@mlops.community Watch it here [https://home.mlops.community/home/videos/exploring-long-context-language-models] [Can Long-context Language Models Subsume Retrieval, RAG, SQL, and More?,](https://arxiv.org/abs/2406.13121) ## Hidden Gems Interested in partnering with us? Get in touch: partners@mlops.community Thanks for reading. See you in Slack [https://go.mlops.community/slack], YouTube [https://www.youtube.com/channel/UCG6qpjVnBTTT8wLGBygANOQ?view_as=subscriber], and podcast [https://home.mlops.community/public/content/] land. Oh yeah, and we are also on X [https://twitter.com/mlopscommunity]. The MLOps Community newsletter is edited by Jessica Rudd [https://www.linkedin.com/in/jmrudd/]. --- Source: https://aaif.live/newsletters/mlopscommunity/2024-08-29-do-we-ask-and-answer-all-the-big-queries-yes-we-do