--- title: "If you build it, they will come..." newsletter: "MLOps Community" date: 2024-05-16 source: https://aaif.live/newsletters/mlopscommunity/2024-05-16-if-you-build-it-they-will-come --- # If you build it, they will come... *Plus, Spotify AI, top tips on efficiency, DREAMs can come true, and hidden gems* *MLOps Community — Agentic AI Foundation, 2024-05-16* ## From A Coding Startup to AI Development in the Enterprise // Ryan Carson // MLOps Podcast #231 Just like authors tell aspiring writers to just write, Ryan tells developers to just start building. Sure, your product might not be perfect at first, but you've got to start somewhere, and you’ll learn as you go. He’s got so much positivity and drive he’d even be able to find positives from our launches of ArcaVacua and USE302 over the last two weeks. This enthusiasm comes through as he talks about his career shift from Treehouse to Intel and shares his goal to make AI development more accessible and the impact it’s had on different sectors. He talks about the value of in-person connections and building quality relationships, even suggesting setting weekly goals to connect with others, which links with the importance of aligning individual efforts with broader company objectives through OKRs and KPIs to make sure you’re having an impact even at a company with thousands of employees. He also introduces Gaudi Three, a new ASIC designed specifically for AI workloads, and shares his positive outlook on future tech advancements and how quick it is to ship a product these days. Just need to get to work on the startup ideas mentioned before others steal them... Video [https://home.mlops.community/home/videos/from-a-coding-startup-to-ai-development-in-the-enterprise] || Spotify [https://open.spotify.com/episode/1urb7hbeOqsCGF1gb7I1Fe?si=ezbdKSdwSEqY-KRMED-IUQ] || Apple [https://podcasts.apple.com/us/podcast/from-a-coding-startup-to-ai-development-in/id1505372978?i=1000655206650] [https://podcasts.apple.com/us/podcast/from-a-coding-startup-to-ai-development-in/id1505372978?i=1000655206650](https://podcasts.apple.com/us/podcast/from-a-coding-startup-to-ai-development-in/id1505372978?i=1000655206650) [https://home.mlops.community/home/videos/from-a-coding-startup-to-ai-development-in-the-enterprise](https://home.mlops.community/home/videos/from-a-coding-startup-to-ai-development-in-the-enterprise) [https://open.spotify.com/episode/1urb7hbeOqsCGF1gb7I1Fe?si=ezbdKSdwSEqY-KRMED-IUQ](https://open.spotify.com/episode/1urb7hbeOqsCGF1gb7I1Fe?si=ezbdKSdwSEqY-KRMED-IUQ) ## Foundational Embeddings for Transfer Learning in Recommender Systems // Sanket Gupta // MLOps Podcast #232 I’m always on the lookout for the next big thing in Gregorian chants, so I’m grateful for Spotify's recommendation systems when it flags up bangers like "Salve Regina" while not pushing any of that modern Znamenny Chant rubbish the kids call music these days. That’s why it was great to have Sanket on to talk about the challenges and methods used in these systems. He discussed focusing on foundational embeddings for transfer learning, their evaluation methods, and their application in search, personalized playlists, and ad targeting. He highlighted the importance of ranking metrics like MRR and NDCG for user engagement and satisfaction. Vector databases and feature stores play a crucial role in managing user preferences and enabling real-time system updates. Sanket also addressed the complexities of balancing system responsiveness with scalability and the continuous efforts to enhance model performance through strategic feature updates and rigorous evaluations. You don’t need AI to recommend a listen to this one! Video [https://home.mlops.community/home/videos/foundational-embeddings-for-transfer-learning-in-recommender-systems] || Spotify [https://open.spotify.com/show/7wZygk3mUUqBaRbBGB1lgh?si=0724ecae41d04b96]|| Apple [https://podcasts.apple.com/us/podcast/mlops-community/id1505372978] [https://podcasts.apple.com/us/podcast/mlops-community/id1505372978](https://podcasts.apple.com/us/podcast/mlops-community/id1505372978) [https://home.mlops.community/home/videos/foundational-embeddings-for-transfer-learning-in-recommender-systems](https://home.mlops.community/home/videos/foundational-embeddings-for-transfer-learning-in-recommender-systems) [https://open.spotify.com/show/7wZygk3mUUqBaRbBGB1lgh?si=0724ecae41d04b96](https://open.spotify.com/show/7wZygk3mUUqBaRbBGB1lgh?si=0724ecae41d04b96) ## Job of the Week [https://ww.wd1.myworkdayjobs.com/en-US/careers/job/Senior-MLOps-Engineer_R240000001196](https://ww.wd1.myworkdayjobs.com/en-US/careers/job/Senior-MLOps-Engineer_R240000001196) ## Making the ML Development Process Mature & Sustainable // Viking Björk Friström & Albin Sundqvist // IRL #77 Stockholm ## MLOps Community IRL Meetup Making the ML Development Process Mature & Sustainable // Viking Björk Friström & Albin Sundqvist // IRL #77 Stockholm Save time and avoid losing your keys by simply storing them in the lock. That’s just one tip to help your life be more efficient and manageable. But what about tips to make your ML development process more efficient and manageable? Well, Viking and Albin have got you covered like a good home insurance policy. They emphasized integrating MLOps early in the process, including establishing model baselines, planning data pipelines, and implementing monitoring from the start. Simplicity should be prioritized for quicker deployments and easier maintenance. Additionally, they highlighted the importance of setting up well-organized repositories for efficient testing and validation. Treating pipelines as artifacts and balancing automation with manual processes can accelerate time to market and consistently deliver value. I suppose the best tip I could give is to watch it. Watch it here [https://home.mlops.community/home/videos/making-the-ml-development-process-mature-and-sustainable] [Watch it here](https://home.mlops.community/home/videos/making-the-ml-development-process-mature-and-sustainable) ## DREAM: Distributed RAG Experimentation Framework Having nightmares trying to sort your RAG set up? This blog introduces the DREAM, a Kubernetes-native architecture designed to optimize RAG setups in a distributed environment. It integrates Ray for distributed computing, LlamaIndex for data indexing, Ragas for synthetic data, MLflow for experiment tracking, MinIO for object storage, Jupyter for interactive computing, and ArgoCD for continuous delivery. This framework enables seamless setup, execution, and monitoring of RAG experiments, ensuring efficient resource utilization and scalability in machine learning operations. DREAM provides a comprehensive solution for managing complex machine learning workflows in a distributed environment. Now there’s no more sleepless nights! With thanks to Aishwarya Prabhat for their contribution. [DREAM: Distributed RAG Experimentation Framework](https://mlops.community/dream-distributed-rag-experimentation-framework/) ## Hidden Gems ## IRL Meetups Scotland [https://www.meetup.com/scotland-mlops-community/events/300633800/] - May 16 San Francisco [https://www.meetup.com/mlops-community/events/300785514/] - May 17 (📣 shoutout to Aporia) Amsterdam [https://www.meetup.com/amsterdam-mlops-community/events/300696760/] - May 28 Munich [https://www.meetup.com/munich-mlops-community/events/300673506/] - May 28 Stockholm [https://www.meetup.com/stockholm-mlops-community/events/300881538/] - May 30 [Scotland](https://www.meetup.com/scotland-mlops-community/events/300633800/) 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/2024-05-16-if-you-build-it-they-will-come