Sultan Ennab

System architecture for agentic AI. What does it take to make AI safe and scalable? This is a collaboration with Anand Nimkar - AI partner @ Deloitte - on enterprise considerations that we have both encountered while building reliable, scalable agentic systems. Start with how you will host models ... SLMs ... LLMs ... model registry ... LLM proxy for external commercial models ... will you be self-hosting GPU compute Establish semantic knowledge store ... vector DB cluster ... kafka, A/V, IOT, even streams of real-time knowledge How will you run multi-agent clusters for scale and resilience? What will be cross-cluster, and cross-agent protocols supported? Establish a memory persistence model (cough, it's implicit with Akka). Will you have private, shared, or external MCP servers? What is their auth? Agentic orchestration will need to be front-ended by an API serving and gateway layer that supports HTTP, gRPC, JSON-RPC, MCP, etc. You'll need to have auto-elasticity, multi-team access controls, and persistence mgmt policies. (cough, Akka Automated Operations handles all this for you). Deployment requires a registry for staging assets within each region. Will you have multi-region operations? Multiple regions to isolate data sovereignty or to execute failover? What is the replication strategy between regions? Akka's memory is active-active replicated across regions with data filters and disaster recovery. Coordinate security (access of systems, and filtering of data) with corporate key value for secrets, AD/IDP provider that controls all inbound / outbound service access. Wire up code repositories to CI/CD. What will be your AI-assisted DevEx. (cough, Akka provides AI-assistance that can generate entire Akka systems through prompts). Evaluation must be integrated inline to agentic clusters (so that agents can make performance, accuracy or cost control flow decisions) but also integrated to 3rd party observability and evaluation vendors. Writing agents is a lot of fun. What's more fun is watching scale, safely in production. System architecture (and Akka and Deloitte) is how you do that.

212

9 Comments

Thanks to NYSE Wired, John Furrier and Brian J. Baumann for the chat at the RAISE Summit on how Stack Overflow is transforming rapidly in the age of AI to become the most vital and trusted source for technologists. We discussed how: 1. Our public platform is changing from a pure Q&A site to a multi-dimensional site that includes multiple features to connect community members, power their learning and unlock their growth in their careers 2. Our private, enterprise product, Stack Overflow Internal, is playing a key role as the trusted, knowledge intelligence layer inside organizations by leveraging our highly effective human curation systems - as companies bring together the power of AI with the expert human in the loop through our platform 3. Our business model has quickly evolved from an Ads / Job listings website to an enterprise and data company https://lnkd.in/gxbTh4WA

107

5 Comments

I recently joined Tony Peleska on the Tech Together podcast hosted by Allison Barmann and Shannon Seaver NBCT to explore what it means to lead and learn in a time of accelerating technological change. We discussed how AI and data are reshaping every profession, why mentorship is essential for resilience, and what it takes to build learning environments that are both future facing and ethically grounded. These are not abstract questions. They are strategic challenges that demand clarity, collaboration, and a commitment to human agency. The conversation moved beyond trends and into the deeper work of preparing students, educators, and professionals to adapt with purpose. I appreciated the opportunity to share practical approaches to lifelong learning, interdisciplinary thinking, and responsible innovation. Thanks to the Tech Together team for creating space where these conversations can unfold with nuance and intention. If you are thinking about how to navigate the future of work with integrity and foresight, this episode offers a meaningful starting point. https://lnkd.in/gAVau9Fd #TechTogetherPodcast #EthicalAI #FutureOfWork #MentorshipMatters #STEMLeadership #LifelongLearning #ResponsibleInnovation

56

1 Comment

The White House just released America’s AI Action Plan that outlines a strategy to accelerate AI leadership, build robust tech infrastructure, and strengthen the country’s global position. It was good to see the degree to which they have lined up top tech executives to help the government formulate a winning strategy. Key highlights: Exporting secure, full-stack American AI to allies Fast-tracking permits for data centers and chip factories Removing barriers to foster innovation and adoption Exciting to see a comprehensive plan to drive U.S. innovation and keep AI progress moving forward Regardless of partisan views one way or another; AI is a place we should want this country to continue to win. It's exciting to see a comprehensive plan to drive U.S. innovation and keep AI progress moving forward #AI #Summit #Government See the full video here https://lnkd.in/eRsDMp-r

34

5 Comments

15 vertical AI investments. 4 winning models. The question isn’t whether you should build in Vertical AI. It’s which model will take you to market leadership. Our thoughts here: https://lnkd.in/gCuz9csW What did we miss? #StartupScaling #AIPlaybooks

48

5 Comments