Reza Rahman

Although big and chunky cloud MPP databases are cool, I've always been a fan of open-source, compact, no-dependencies software, where vibrant community support and genuine love for product (and users) allow it to thrive among many commercial counterparts. #SQLite is great, so is #PostgreSQL, and now #DuckDB is emerging as the "hummingbird" of OLAP databases - small, fast and easy to manage. While it’s not (yet) as mature or feature-rich as some entrenched incumbents, it has already shown an immense potential for middleware-type workloads, where speed, #SQL expressiveness and interoperability are key. "AWS S3 data ingestion and augmentation patterns using DuckDB and Python" @

5

1 Comment

A few thoughts on #LLM usage in the small, prompted by my toying with #llama3 and #phi-3 on consumer hardware.

5

94% accuracy across 11 types of cancer! The way the world is running today will be drastically different in the near future. To fully realize AI, having the AI-ready data foundation is not enough, we also need the AI-led business process revolution! #AI #BusinessProcess

67

2 Comments

Amazon Web Services (AWS) rolled out v1.0 of cost data export compliant with FinOps Foundation FOCUS specifications. zOpt.ai is a contributor to #FinOps #FOCUS specifications and I am glad to see the progress and adoption of the standard specifications by all the major cloud providers. This is a major step in understanding cloud spend across cloud providers in uniform manner.

23

RAG architectures enhance #LLMs by integrating proprietary data, but maintaining data freshness poses challenges. 🤯 Continuous updates of vector embeddings ensure accuracy in AI models’ responses. ✅ The below solution uses MongoDB Atlas Stream Processing and MongoDB Atlas Vector Search to streamline the updating and searching of embeddings, simplifying the process for developers. 👇 https://lnkd.in/gS87PqEt

1

Mixture-of-Agents (MoA) methodology is a new approach that leverages the collective strengths of multiple LLMs. In this is construct a layered MoA architecture wherein each layer comprises multiple LLM agents. Each agent takes all the outputs from agents in the previous layer as auxiliary information in generating its response. MoA models achieves state-of-art performance on AlpacaEval 2.0, MT-Bench and FLASK, surpassing GPT-4 Omni. For example, our MoA using only open-source LLMs is the leader of AlpacaEval 2.0 by a substantial gap, achieving a score of 65.1% compared to 57.5% by GPT-4 Omni. Paper details here: https://lnkd.in/d6eAmtww

5

Palantir is what's next in software. It's the OS for the cloud. The rest of the world is catching on, slowly. I look froward to the day Forrester recognizes what a cloud OS is, and how Palantir is 1:1 in this space.

59

3 Comments

AWS re:Invent 2024 delivered groundbreaking announcements with significant implications for cloud adoption in the public sector. Here's my perspective as a Chief Architect at Accenture Federal Services: Generative AI Takes Center Stage Amazon Bedrock: Enhanced features like model distillation enable agencies to deploy powerful AI solutions even with limited resources or strict security requirements. Bedrock's focus on data security and privacy is crucial for compliance with federal regulations. Amazon Nova: These new foundation models can revolutionize how agencies process and understand data, opening doors for innovation in areas like natural language processing and synthetic data generation. Trainium2 UltraServers: These servers accelerate complex AI workloads while adhering to stringent security and compliance standards, enhancing the performance of mission-critical applications. Compute Gets a Performance Boost Amazon EC2 Trn2 Instances: These instances offer attractive price-performance benefits for budget-conscious agencies. Thorough evaluation of their security posture is crucial before widespread adoption. AWS Trainium3: This signals AWS's commitment to purpose-built AI acceleration, ensuring federal agencies have continued access to cutting-edge technology. Amazon EC2 P5en Instances: With high bandwidth and networking capabilities, these instances are ideal for demanding federal workloads like high-performance computing. Integration with AWS GovCloud (US) will be key for supporting sensitive applications. Storage and Database Advancements Amazon S3 Innovations: New features like S3 Tables and S3 Metadata offer valuable tools for data management. Implementation must align with federal requirements for data governance and compliance. Amazon Aurora DSQL: This capability offers possibilities for modernizing legacy applications by bridging the gap between relational and document data. Looking Ahead re:Invent 2024 showcased the transformative potential of cloud computing for the public sector. Accenture Federal Services is committed to helping our clients navigate this landscape and leverage these innovations to achieve their mission objectives. https://lnkd.in/eF6jTVsg #AWS #reInvent #GenerativeAI #CloudComputing #FedGov #Innovation #AccentureFederalServices

75

1 Comment

Great to see the release of this new model Phi-4 from Microsoft Research and how a small language model can often out-perform an LLM https://lnkd.in/eRPZhkVZ

15

Continuous learning is crucial to retaining IT talent. CIOs who report low skill gaps in their organization are focused on upskilling/reskilling IT staff and hiring internally. Build an internal talent pipeline effectively. Ask me how Gartner can help you identify and fix the developer experience problems that cause retention issues. #GartnerIT #MidsizeEnterprise #Talent

7

5 Comments

Want your AI applications to be more reliable? Don't let bad APIs hold your AI back. Learn how to bridge the gap between your data and AI applications with robust, high-quality APIs, with Apigee API Management and Mabl. Register for our free webinar and discover how to build consistent APIs that deliver reliable results. In this session, you will learn how - Apigee helps you to build and manage APIs efficiently - Mabl helps ensure reliability for your APIs

1

Enabling use cases like factory floor optimization, assisted workforce, and asset tracking can be expensive and time consuming. Building, deploying, and maintaining configuration can add to the complexity. In this webinar, you will learn from the experts at Google Cloud and ClearObject how to leverage the very latest in AI, modern infrastructure, and edge computing to enable modern manufacturing outcomes in your center today.

1

Incredibly proud of the latest Kinesis Client Library version (3.0) which, when compared to previous KCL versions, reduces compute costs by up to 33%. KCL 3.0 introduces an enhanced load balancing algorithm that continuously monitors resource utilization of the stream processing workers and automatically redistributes the load from over-utilized workers to other underutilized workers. This ensures even CPU utilization across workers and removes the need to over-provision the stream processing compute workers which reduces cost. Additionally, KCL 3.0 is built with the AWS SDK for Java 2.x for improved performance and security features, fully removing the dependency on the AWS SDK for Java 1.x. #Kinesis #DataStreaming #KCL

94

I have been watching for this move for long would be interesting to see how the PTU model will compare with the GPT’s https://lnkd.in/eW8xwSEP

1

In the video "𝐏𝐥𝐚𝐭𝐟𝐨𝐫𝐦 𝐚𝐬 𝐜𝐨𝐝𝐞: simplifying developer platform design with reference architectures," Stephan Schneider and Mike Gatto from McKinsey & Company address developers' complexities and challenges in enterprise environments, proposing robust solutions through standardized developer platforms. Stefan opens the discussion by outlining the current hurdles in enterprise settings, such as lack of standardization, legacy technology burdens, and inefficient toolchains, all contributing to a diminished developer experience and innovation slowdown. He emphasizes the need for streamlined engineering processes to enhance productivity and speed up the development lifecycle, from coding to deployment. Stefan's approach involves removing unnecessary complexities and empowering developers with self-service capabilities. It enables them to focus more on coding and less on operational logistics, reducing the cognitive load and manual tasks hindering developer efficiency. Mike follows up by diving into the technical aspects of implementing an MVP developer platform using cloud-native reference architectures. He details the process starting from version control, where developers check their code, through CI pipelines that build and deploy these applications into containers within Kubernetes environments. Mike showcases a live demo to illustrate the deployment of a three-tier banking application on AWS, using a combination of a workload specification tool and a platform orchestrator to separate developer and operations concerns effectively. Mike's part of the talk mainly focuses on how these architectures facilitate rapid deployment and ensure compliance with security standards while allowing flexibility across cloud environments like AWS, Azure, and GCP. His explanation of the technical workflow demonstrates the practical application of these concepts, providing a clear, actionable framework for organizations looking to adopt similar strategies. 𝐂𝐨𝐧𝐜𝐥𝐮𝐬𝐢𝐨𝐧 The video highlights the challenges faced by developers in large enterprises due to outdated practices, complex legacy systems, and manual tasks that impact productivity. The proposed solution is to standardize development practices and empower developers through automated, self-service platforms. Mike explains the MVP developer platform tools to implement these solutions in a cloud environment, providing practical insight into enhancing developer experiences and productivity. This note summarizes the video's content and suggests watching it to thoroughly understand complex technologies, especially in software development and DevOps. The video demonstrates the implementation and functioning of complex technologies and motivates viewers to explore similar solutions. https://lnkd.in/ezHuzvdf #platformengineering #devops #softwareengineering #softwarearchitecture #productivity #innovation #standardization #complexity #cloudnative #kubernetes #aws #azure #gcp

2

Some interesting thoughts from Microsoft CEO Satya Nadella on AI agents becoming the new backend logic, making SaaS applications "obsolete". This evolution reduces the need for predefined logic in application code, allowing AI agents to interact directly with databases and APIs, adapt to new scenarios, and perform operations across multiple repositories.

9

1 Comment