--- title: "Before you power off and get the concrete ready..." newsletter: "MLOps Community" date: 2024-08-08 source: https://aaif.live/newsletters/mlopscommunity/2024-08-08-before-you-power-off-and-get-the-concrete-ready --- # Before you power off and get the concrete ready... *Plus, the ultimate balancing act: speed and safety, just because you're irresponsible doesn't mean your AI has to be, be observant, ML Kickstarter, and hidden gems.* *MLOps Community — Agentic AI Foundation, 2024-08-08* https://www.linkedin.com/posts/pramod-gosavi-b32a71_google-hires-characterai-cofounders-and-activity-7225202718397677569-xIFc/ ## Harnessing AI APIs for Safer, Accurate, & Reliable Applications // Ron Heichman //MLOps Podcast #252 Harnessing AI APIs for Safer, Accurate, & Reliable Applications // Ron Heichman //MLOps Podcast #252 Without following Gene Spafford’s quote about secure systems, what can we do? In this episode with Ron, we discuss privacy and compliance issues, especially in sensitive fields like banking. He talks about integrating LLMs into specialized industry solutions, contrasting this with traditional machine learning. We also explore how LLMs can be incorporated into products like HR software and customer support, the challenges in generating diverse training data, and strategies for prompt engineering and preventing jailbreaks. At least now, you won't have to get the concrete ready. Video [https://home.mlops.community/home/videos/harnessing-ai-apis-for-safer-accurate-and-reliable-applications] || Spotify [https://open.spotify.com/episode/7pTwCe2Wf8LHHcImFfZr82?si=GVRY_1_mQc2Diuj3Vi9AdA] || Apple [https://podcasts.apple.com/us/podcast/harnessing-ai-apis-for-safer-accurate-reliable-applications/id1505372978?i=1000664498651] [https://podcasts.apple.com/us/podcast/harnessing-ai-apis-for-safer-accurate-reliable-applications/id1505372978?i=1000664498651](https://podcasts.apple.com/us/podcast/harnessing-ai-apis-for-safer-accurate-reliable-applications/id1505372978?i=1000664498651) [https://home.mlops.community/home/videos/harnessing-ai-apis-for-safer-accurate-and-reliable-applications](https://home.mlops.community/home/videos/harnessing-ai-apis-for-safer-accurate-and-reliable-applications) [https://open.spotify.com/episode/7pTwCe2Wf8LHHcImFfZr82?si=GVRY_1_mQc2Diuj3Vi9AdA](https://open.spotify.com/episode/7pTwCe2Wf8LHHcImFfZr82?si=GVRY_1_mQc2Diuj3Vi9AdA) ## MLOps Community Reading Group Don't forget, our new reading group starts today! Join us 11:00 - 12:00 ET when we'll be discussing: Can Long-Context Language Models Subsume Retrieval, RAG, SQL [https://arxiv.org/pdf/2406.13121] Register here [https://lu.ma/x2nf6mn4], and join the Slack channel #reading-group [https://mlops-community.slack.com/archives/C01R8Q26UE4]. Look forward to seeing you there! [Can Long-Context Language Models Subsume Retrieval, RAG, SQL](https://arxiv.org/pdf/2406.13121) ## Balancing Speed and Safety // Panel // AIQCON ## AIQCON Panel I feel the need—the need for safety! Sure, you can deploy fast, but can you do it safely too? This panel focused on balancing rapid deployment with safety, highlighting the context-dependent nature of risks and the necessity of thorough evaluation. Strategies included involving diverse teams, such as legal and risk management, for comprehensive oversight and accountability. The chat consistently emphasized setting clear success criteria and understanding model limitations. Give it a watch, and we'll be your wingman anytime. Video [https://home.mlops.community/home/videos/balancing-speed-and-safety] || Spotify [https://open.spotify.com/episode/1IWsoWo6zeU6VyXwqo4t12?si=L0bpCatqTnWKhO1cF0mcUw] || Apple [https://podcasts.apple.com/us/podcast/balancing-speed-and-safety-panel-aiqcon/id1505372978?i=1000664153008] ## Job of the Week [https://www.linkedin.com/jobs/view/3913790033/](https://www.linkedin.com/jobs/view/3913790033/) ## Responsible AI in Action // Panel // MLOps Community IRL Meetup #89 Lisbon ## MLOps Community IRL Meetup I wonder if AI will ever have a 'Use Responsibly' caution, like we see with alcohol. I imagine it’ll get ignored just as much. Until then, we've got this IRL panel that discussed responsible AI, fairness, transparency, security, and ethics. Continuous learning was deemed essential, prioritizing mandatory training and self-directed learning to stay updated with the latest technologies and best practices. Practical takeaways were highlighted, such as maintaining clear communication with stakeholders throughout the project lifecycle to manage expectations and align with business goals. Additionally, balancing innovation with reliability and explainability when using LLMs, and incorporating end-user feedback during product development were considered crucial for enhancing usability and trust in AI solutions. I'm off to down a bottle of Everclear and put in prompts about drunk AI decisions. Watch it here [https://home.mlops.community/home/videos/responsible-ai-in-action] [Watch it here](https://home.mlops.community/home/videos/responsible-ai-in-action) ## MLOps Coding Course: Mastering Observability for Reliable MLThey say seeing is believing. But observing is knowing. Following up Part 1, this blog examines the tools and practices for achieving comprehensive observability in your ML projects. It unravels key concepts like proactive monitoring to track data and model drifts, setting up alerts with tools like Datadog, and employing explainability techniques using SHAP. Plus, it showcases practical code examples from the accompanying MLOps Python Package, and explores the benefits of integrating industry-leading solutions like MLflow for model monitoring, data lineage tracking, and performance optimization. I won’t be able to observe you reading it, but I know you will.With thanks to Médéric Hurier for their contribution. Introducing the ML Kickstarter You’ve got an idea for a personal ML project to predict who’s next to leave OpenAI, but these things are time-consuming to build. Helpfully, this blog introduces an efficient project structure to simplify the process. The system includes essential features like model experiment tracking, a model registry, a feature store, data validation, versioning, pipeline orchestration, and model serving. Using Docker for environment setup and Prefect for workflow management, the project ensures a streamlined development experience without relying on cloud providers. Key tools such as mlflow, aligned, streamlit, and ollama are integrated to create a robust, cost-effective, and developer-friendly ML framework that supports end-to-end functionality on a local machine. Now you’ll know which board position to apply for before they do. With thanks to Mats Eikeland for their contribution. [MLOps Coding Course: Mastering Observability for Reliable ML](https://home.mlops.community/home/blogs/mlops-coding-course-mastering-observability-for-reliable-ml) ## Hidden Gems ## Summer Season 🌴 ## IRL Meetups Looks like everyone's away on holiday! But you can still find your local chapter here [https://mlops.community/meetups/] and stay up to date with their events. If there isn't one local enough and you want to start your own local chapter, reach out to us: binoy@mlops.community [here](https://mlops.community/meetups/) 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/2024-08-08-before-you-power-off-and-get-the-concrete-ready