Kubernetes not best for large-scale Python workloads: Coiled founder

Data scientist Matthew Rocklin argues Kubernetes isn't the best way to run large-scale Python workloads in the cloud. His company, Coiled, offers an alternative. By David Cassel

Runhouse team has been on a tear 🚀 We soft launched this at PyTorch Conference last week and are officially announcing Open Source Kubetorch today. Between OSS and a serverless option (in the works, let's chat), it's easier than ever to scale your ML workloads on Kubernetes. Write simple Python to define compute (GPU/CPU, memory, image) and dispatch to your cluster on any cloud - no YAML or Dockerfiles needed. https://lnkd.in/g_8jq_BJ

Excited to share my latest project — AI Voice & Text Chatbot using Groq API, FastAPI & Redis! This intelligent chatbot can understand both voice and text queries, generate real-time responses using Groq’s LLMs, and handle contextual memory via Redis Cloud. ⚙️ Tech Stack: FastAPI Groq API (LLaMA 3.1) Redis Python SpeechRecognition gTTS 💡 Key Features: • Voice-to-text and text-to-voice communication • Real-time AI conversation using Groq API • FastAPI backend with Redis caching • Scalable and deployable as a web-based assistant 🔗 GitHub Repository: https://lnkd.in/dW9fNr2v 📊 Architecture Overview is attached below. I learned a lot about integrating APIs, handling real-time speech, and optimizing latency through caching. Would love to hear your thoughts and suggestions! 🚀 #AI #FastAPI #Groq #Redis #VoiceChatbot #Python #LLM #MachineLearning #SoftwareDevelopment

Want to run Python code in the cloud for free? This edition of my article explores its key features, how to use it effectively, and why it’s a game-changer for data science and machine learning projects. Read more inside!

🚀 Automate your AWS Cloud Costs with Python + Serverless = FinOps Superpower Hi network! As a Cloud Engineer passionate about automation, I wanted to share a pattern I came across, for FinOps optimisation using AWS Lambda + Python to drive cost savings and better visibility across the cloud. What the automation does: • Pulls cost & usage data (e.g., from AWS Cost and Usage Report) and processes it with Python (via Pandas, Boto3) to identify idle/over-provisioned resources. • Triggers Lambda functions on schedule or event (e.g., new CUR file arrival, or anomaly detected) to act or alert. • Integrates tagging, budgets, anomaly detection & alerts for governance (using guidance from Amazon Web Services’s Cloud Financial Management white-paper). Why this matters: • Cost = innovation blocker. Too many idle/unused resources or inefficient pricing (e.g., On-Demand when Spot/Reserved would suffice) raise spend. • Automation means faster decision making (no manual CSV-sifting), more consistent FinOps practices and less human overhead. • With serverless (Lambda) you pay only for what you execute—fits FinOps mindset of “cost aligned to value”. My key take-aways & best practices: Establish strong tagging (business unit, cost centre, environment) before heavy analysis. (See AWS CFM guidance) Build your pipeline: ingestion → analysis → alert/action → dashboard. Use Python for the “brains” (e.g., detect idle EC2s, identify unused RDS, check Lambda memory over-provisioning) and Lambda for the “execution engine”. Visualise trends and anomalies: dashboards that show month-on-month cost change, idle resource counts, savings potential. Iterate — once you automate one cost-leakage use-case, roll out new ones (e.g., Spot-instance recommendation, S3 lifecycle enforcement). If you’re working in cloud cost management or FinOps, I’d love to connect + swap ideas on automations that drive real savings. #AWS #Serverless #FinOps #CloudAutomation #Python #AWSCostOptimization #CloudEngineering

🚀 Architecting the Future: Launching into Python for Serverless & Automation on AWS! Starting a strategic course focused on mastering the intersection of Python and advanced AWS services. This isn't just about scripting; it’s about building highly scalable, secure, and automated cloud solutions. The Core Pillars of Cloud Mastery The curriculum is meticulously designed to move from robust Python fundamentals (covering variables, conditionals, functions, lists, and dictionaries) to sophisticated enterprise architecture: Serverless Architectures: I'll be deploying complete serverless solutions using Amazon API Gateway, AWS Lambda, and Amazon DynamoDB. This foundation is critical for developing modern, cost-efficient microservices. Intelligent Automation: The focus then shifts to practical, real-world application, writing and running powerful automation scripts with Python and AWS Systems Manager. This empowers essential automation for Support and Operations. Data & Security: The course also integrates crucial topics like managing data with Amazon RDS and Amazon DynamoDB , all while enforcing robust access control using AWS IAM roles and safeguarding credentials with AWS Secrets Manager. This course is designed to give a deep conceptual understanding of Python and how to use it strategically for enterprise-level automation. I am committed to leveraging these skills to drive significant efficiency and innovation in cloud operations. Let's Connect on Cloud Strategy! For my network focused on cloud and DevOps: What Python library or AWS service do you believe is currently the single most underestimated tool for automation? Share your insights below—I'm ready to learn from your experience! 👇 #AWS #Python #Serverless #Automation #CloudArchitecture #LinkedInLearning #CareerGrowth

Unleash the full power of your data science projects with my latest Medium article: “From Cloud Notebooks to Enterprise Deployment”. I walk through how to take your exploratory notebooks—whether in Colab, Jupyter or Google Cloud—and transform them into scalable, production‑ready enterprise solutions. Learn how to build reliable pipelines, integrate with enterprise systems, and deliver real business value beyond the prototype phase. Check out the full article here: https://lnkd.in/gEzmKn56 #DataScience #MLOps #CloudNotebooks #EnterpriseAI #AnalyticsTransformation #MLDeployment #BigData #MachineLearning #TechLeadership

Today, AgentCore Runtime has launched a second method to deploy agents – direct code deployment (for Python). Agent code and its dependencies can be packaged as a zip archive, alleviating the need for Docker definition and ECR dependencies. This makes it straightforward for developers to prototype and iterate faster. This method is a good fit for customers who prefer not to worry about Docker expertise and container infrastructure when deploying agents. https://lnkd.in/dtYYt-fh #AWS #AmazonWebServices #AWSBlogs #Cloud #CloudComputing #GenAI #BedRock #MachineLearning

This week, I deep dived into something that quietly saves huge amounts of cost and time data compression in GCP. I wanted to see how GZIP and Snappy behave across different layers like BigQuery, Cloud Storage, and Dataflow pipelines. The goal? - Find the right balance between storage efficiency and processing speed. Here’s what I found: 🔹 BigQuery automatically handles compression using its columnar storage format you don’t have to lift a finger. 🔹 Parquet + Snappy turned out to be ideal for analytical pipelines. 🔹 GZIP gave better compression ratios, making it great for archiving and long-term storage. 🔹 Smart compression choices directly reduced GCS storage size and improved pipeline execution times. I also created a small .ipynb notebook on GitHub showing how to test and visualize compression results feel free to check it out! https://lnkd.in/dcKDCPP8 Here’s a quick snapshot from the experiment #DataEngineering #GoogleCloud #BigQuery #DataCompression #Python #ETL #GCP #DataPipeline #CloudComputing

Proper logging is undoubtedly a game-changer when you're trying to track down bugs or get to the root cause of issues in your application. Setting this up in Python is actually pretty straightforward. Here is a blog post I recently wrote walking through how to configure logging that sends your logs directly to CloudWatch which is a service that monitors applications. https://lnkd.in/gkUp2T5m