Agentic Conversations (formally mlops.community)

Kris Beevers is the CEO at NetBox Labs, working on turning NetBox into the system of record and automation backbone for modern and AI-driven infrastructure.

Speed and Scale: How Today's AI Datacenters Are Operating Through Hypergrowth // MLOps Podcast #359 with Kris Beevers, CEO of NetBox Labs

Join the Community: https://go.mlops.community/YTJoinIn

Get the newsletter: https://go.mlops.community/YTNewsletter

MLOps GPU Guide: https://go.mlops.community/gpuguide

// Abstract

Hundreds of neocloud operators and "AI Factory" builders have emerged to serve the insatiable demand for AI infrastructure. These teams are compressing the design, build, deploy, operate, scale cycle of their infrastructures down to months, while managing massive footprints with lean teams. How? By applying modern intent-driven infrastructure automation principles to greenfield deployments. We'll explore how these teams carry design intent through to production, and how operating and automating around consistent infrastructure data is compressing "time to first train".

// Bio

Kris Beevers is the Co-founder and CEO of NetBox Labs. NetBox is used by nearly every Neocloud and AI datacenter to manage their networks and infrastructure. Kris is an engineer at heart and by background, and loves the leverage infrastructure innovation creates to accelerate technology and empower engineers to do their best work. A serial entrepreneur, Kris has founded and helped lead multiple other successful businesses in the internet and network infrastructure. Most recently, he co-founded and led NS1, which was acquired by IBM in 2023. He holds a Ph.D. in Computer Science from Rensselaer Polytechnic Institute and is based in New Jersey.

// Related Links

Website: https://netboxlabs.com/

Coding Agents Conference: https://luma.com/codingagents

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

Join our Slack community [https://go.mlops.community/slack]

Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

Sign up for the next meetup: [https://go.mlops.community/register]

MLOps Swag/Merch: [https://shop.mlops.community/]

Connect with Demetrios on LinkedIn: /dpbrinkm

Connect with Kris on LinkedIn: /beevek/

Timestamps:

[00:00] Observability and Delta Analysis

[00:26] New World Exploration

[04:06] Bottlenecks in AI Infrastructure

[13:37] Data Center Optimization Challenges

[19:58] Tech Stack Breakdown

[25:26] Data Center Design Principles

[31:32] Constraints and Automation in Design

[40:00] Complexity in Data Centers

[45:02] GPU Cloud Landscape

[50:24] Data Centers in Containers

[57:45] Observability Beyond Software

[1:04:43] Tighter Integrations vs NetBox

[1:06:47] Wrap up

Mike Oaten is the Founder and CEO of TIKOS, working on building AI assurance, explainability, and trustworthy AI infrastructure, helping organizations test, monitor, and govern AI models and systems to make them transparent, fair, robust, and compliant with emerging regulations.

Cracking the Black Box: Real-Time Neuron Monitoring & Causality Traces // MLOps Podcast #358 with Mike Oaten, Founder and CEO of TIKOS

Join the Community: https://go.mlops.community/YTJoinIn

Get the newsletter: https://go.mlops.community/YTNewsletter

// Abstract

As AI models move into high-stakes environments like Defence and Financial Services, standard input/output testing, evals, and monitoring are becoming dangerously insufficient. To achieve true compliance, MLOps teams need to access and analyse the internal reasoning of their models to achieve compliance with the EU AI Act, NIST AI RMF, and other requirements.

In this session, Mike introduces the company's patent-pending AI assurance technology that moves beyond statistical proxies. He will break down the architecture of the Synapses Logger, a patent-pending technology that embeds directly into the neural activation flow to capture weights, activations, and activation paths in real-time.

// Bio

Mike Oaten serves as the CEO of TIKOS, leading the company’s mission to progress trustworthy AI through unique, high-performance AI model assurance technology. A seasoned technical and data entrepreneur, Mike brings experience from successfully co-founding and exiting two previous data science startups: Riskopy Inc. (acquired by Nasdaq-listed Coupa Software in 2017) and Regulation Technologies Limited (acquired by mnAi Data Solutions in 2022).

Mike's expertise spans data, analytics, and ML product and governance leadership. At TIKOS, Mike leads a VC-backed team developing technology to test and monitor deep-learning models in high-stakes environments, such as defence and financial services, so they comply with the stringent new laws and regulations.

// Related Links

Website: https://tikos.tech/

LLM guardrails: https://medium.com/tikos-tech/your-llm-output-is-confidently-wrong-heres-how-to-fix-it-08194fdf92b9

Model Bias: https://medium.com/tikos-tech/from-hints-to-hard-evidence-finally-how-to-find-and-fix-model-bias-in-dnns-2553b072fd83

Model Robustness: https://medium.com/tikos-tech/tikos-spots-neural-network-weaknesses-before-they-fail-the-iris-dataset-b079265c04da

GPU Optimisation: https://medium.com/tikos-tech/400x-performance-a-lightweight-open-source-python-cuda-utility-to-break-vram-barriers-d545e5b6492f

Hyperbolic GPU Cloud: app.hyperbolic.ai.

Coding Agents Conference: https://luma.com/codingagents

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

Join our Slack community [https://go.mlops.community/slack]

Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

Sign up for the next meetup: [https://go.mlops.community/register]

MLOps Swag/Merch: [https://shop.mlops.community/]

Connect with Demetrios on LinkedIn: /dpbrinkm

Connect with Mike on LinkedIn: /mike-oaten/

Timestamps:

[00:00] Regulations as Opportunity

[00:25] Regulation Compliance Fun

[02:49] AI Act Layers Explained

[05:19] Observability in Systems vs ML

[09:05] Risk Transfer in AI

[11:26] LLMs and Model Approval

[14:53] LLMs in Finance

[17:17] Hyperbolic GPU Cloud Ad

[18:16] Stakeholder Alignment and Tech

[22:20] AI in Regulated Environments

[28:55] Autonomous Boat Regulations

[34:20] Data Compliance Mapping

[39:11] Data Capture Strategy

[41:13] EU AI Act Insights

[44:52] Wrap up

[45:45] Join the Coding Agents Conference!

Paulo Vasconcellos is the Principal Data Scientist for Generative AI Products at Hotmart, working on AI-powered creator and learning experiences, including intelligent tutoring, content automation, and multilingual localization at scale.

Join us at Coding Agents: The AI Driven Developer Conference - https://luma.com/codingagents

MLOps GPU Guide: ⁠https://go.mlops.community/gpuguide

Join the Community:

https://go.mlops.community/YTJoinIn

Get the newsletter: https://go.mlops.community/YTNewsletter

// Abstract

“Agent as a product” sounds like hype, until Hotmart turns creators’ content into AI businesses that actually work.

// Bio

Paulo Vasconcellos is the Principal Data Scientist for Generative AI Products at Hotmart, where he leads efforts in applied AI, machine learning, and generative technologies to power intelligent experiences for creators and learners. He holds an MSc in Computer Science with a focus on artificial intelligence and is also a co-founder of Data Hackers, a prominent data science and AI community in Brazil. Paulo regularly speaks and publishes on topics spanning data science, ML infrastructure, and AI innovation.

// Related LinksWebsite: paulovasconcellos.com.br

Coding Agent - Virtual Conference: https://home.mlops.community/home/events/coding-agents-virtual

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

Join our Slack community [https://go.mlops.community/slack]

Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

Sign up for the next meetup: [https://go.mlops.community/register]

MLOps Swag/Merch: [https://shop.mlops.community/]

MLOps GPU Guide: https://go.mlops.community/gpuguide

Connect with Demetrios on LinkedIn: /dpbrinkm

Connect with Paulo on LinkedIn: /paulovasconcellos/

Timestamps:

[00:00] Hotmart Data Science Challenges

[02:38] LLMs vs spaCy

[11:38] Use Cases in Production

[19:04] Coding Agents Virtual Conference Announcement!

[29:27] ML to AI Product Shift

[34:49] Tool-Augmented Agent Approach

[38:28] MLOps GPU Guide

[41:24] AI Use Cases at Hotmart

[49:34] Agent Tool Access Explained

[51:04] MLOps Community Gratitude

[53:22] Wrap up

Wilder Lopes is the CEO and Founder of Ogre.run, working on AI-driven dependency resolution and reproducible code execution across environments.How Universal Resource Management Transforms AI Infrastructure Economics // MLOps Podcast #357 with Wilder Lopes, CEO / Founder of Ogre.runJoin the Community:

https://go.mlops.community/YTJoinInGet the newsletter: https://go.mlops.community/YTNewsletter

// AbstractEnterprise organizations face a critical paradox in AI deployment: while 52% struggle to access needed GPU resources with 6-12 month waitlists, 83% of existing CPU capacity sits idle. This talk introduces an approach to AI infrastructure optimization through universal resource management that reshapes applications to run efficiently on any available hardware—CPUs, GPUs, or accelerators.We explore how code reshaping technology can unlock the untapped potential of enterprise computing infrastructure, enabling organizations to serve 2-3x more workloads while dramatically reducing dependency on scarce GPU resources. The presentation demonstrates why CPUs often outperform GPUs for memory-intensive AI workloads, offering superior cost-effectiveness and immediate availability without architectural complexity.// BioWilder Lopes is a second-time founder, developer, and research engineer focused on building practical infrastructure for developers. He is currently building Ogre.run, an AI agent designed to solve code reproducibility.Ogre enables developers to package source code into fully reproducible environments in seconds. Unlike traditional tools that require extensive manual setup, Ogre uses AI to analyze codebases and automatically generate the artifacts needed to make code run reliably on any machine. The result is faster development workflows and applications that work out of the box, anywhere.// Related LinksWebsite: https://ogre.runhttps://lopes.aihttps://substack.com/@wilderlopes https://youtu.be/YCWkUub5x8c?si=7RPKqRhu0Uf9LTql

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExploreJoin our Slack community [https://go.mlops.community/slack]Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)] Sign up for the next meetup: [https://go.mlops.community/register]MLOps Swag/Merch: [https://shop.mlops.community/]Connect with Demetrios on LinkedIn: /dpbrinkmConnect with Wilder on LinkedIn: /wilderlopes/Timestamps:[00:00] Secondhand Data Centers Challenges[00:27] AI Hardware Optimization Debate[03:40] LLMs on Older Hardware[07:15] CXL Tradeoffs[12:04] LLM on CPU Constraints[17:07] Leveraging Existing Hardware[22:31] Inference Chips Overview[27:57] Fundamental Innovation in AI[30:22] GPU CPU Combinations[40:19] AI Hardware Challenges[43:21] AI Perception Divide[47:25] Wrap up

Corey Zumar is a Product Manager at Databricks, working on MLflow and LLM evaluation, tracing, and lifecycle tooling for generative AI.

Jules Damji is a Lead Developer Advocate at Databricks, working on Spark, lakehouse technologies, and developer education across the data and AI community.

Danny Chiao is an Engineering Leader at Databricks, working on data and AI observability, quality, and production-grade governance for ML and agent systems.

MLflow Leading Open Source // MLOps Podcast #356 with Databricks' Corey Zumar, Jules Damji, and Danny Chiao

Join the Community:

https://go.mlops.community/YTJoinIn

Get the newsletter: https://go.mlops.community/YTNewsletter

Shoutout to Databricks for powering this MLOps Podcast episode.

// Abstract

MLflow isn’t just for data scientists anymore—and pretending it is is holding teams back. Corey Zumar, Jules Damji, and Danny Chiao break down how MLflow is being rebuilt for GenAI, agents, and real production systems where evals are messy, memory is risky, and governance actually matters. The takeaway: if your AI stack treats agents like fancy chatbots or splits ML and software tooling, you’re already behind.

// Bio

Corey Zumar

Corey has been working as a Software Engineer at Databricks for the last 4 years and has been an active contributor to and maintainer of MLflow since its first release.

Jules Damji

Jules is a developer advocate at Databricks Inc., an MLflow and Apache Spark™ contributor, and Learning Spark, 2nd Edition coauthor. He is a hands-on developer with over 25 years of experience. He has worked at leading companies, such as Sun Microsystems, Netscape, @Home, Opsware/LoudCloud, VeriSign, ProQuest, Hortonworks, Anyscale, and Databricks, building large-scale distributed systems. He holds a B.Sc. and M.Sc. in computer science (from Oregon State University and Cal State, Chico, respectively) and an MA in political advocacy and communication (from Johns Hopkins University)

Danny Chiao

Danny is an engineering lead at Databricks, leading efforts around data observability (quality, data classification). Previously, Danny led efforts at Tecton (+ Feast, an open source feature store) and Google to build ML infrastructure and large-scale ML-powered features. Danny holds a Bachelor’s Degree in Computer Science from MIT.

// Related Links

Website: https://mlflow.org/

https://www.databricks.com/

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

Join our Slack community [https://go.mlops.community/slack]Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

Sign up for the next meetup: [https://go.mlops.community/register]

MLOps Swag/Merch: [https://shop.mlops.community/]

Connect with Demetrios on LinkedIn: /dpbrinkm

Connect with Corey on LinkedIn: /corey-zumar/

Connect with Jules on LinkedIn: /dmatrix/

Connect with Danny on LinkedIn: /danny-chiao/

Timestamps:

[00:00] MLflow Open Source Focus

[00:49] MLflow Agents in Production

[00:00] AI UX Design Patterns

[12:19] Context Management in Chat

[19:24] Human Feedback in MLflow

[24:37] Prompt Entropy and Optimization

[30:55] Evolving MLFlow Personas

[36:27] Persona Expansion vs Separation

[47:27] Product Ecosystem Design

[54:03] PII vs Business Sensitivity

[57:51] Wrap up

Euro Beinat is the Global Head of AI and Data Science at Prosus Group, working on scaling AI-driven tools and agent-based systems across Prosus’s global portfolio, deploying internal assistants like Toqan and generative AI platforms such as PlusOne, and building initiatives like AI House Amsterdam and interdisciplinary AI residencies to explore intent-driven AI and strengthen Europe’s AI ecosystem.

Mert Öztekin is the Chief Technology Officer at Just Eat Takeaway.com, working on advancing the company’s platform with AI-driven ordering and personalised user experiences, scaling cloud and generative AI tooling for engineering productivity, and exploring innovative delivery technologies like automation to make ordering and delivery more seamless.

Join the Community: https://go.mlops.community/YTJoinIn

Get the newsletter: https://go.mlops.community/YTNewsletter

MLOps GPU Guide: https://go.mlops.community/gpuguide

// Abstract

Agents sound smart until millions of users show up. A real talk on tools, UX, and why autonomy is overrated.

// Bio

Euro Beinat

Euro is a technology executive and entrepreneur specializing in data science, machine learning, and AI. He works with global corporations and startups to build data- and ML-driven products and businesses. His current focus is on Generative AI and the use of AI as a tool for invention and innovation.

Mert Öztekin

Mert is the current Chief Technology Officer at Just Eat Takeaway.com with previous experience as a CTO at Delivery Hero Germany GmbH, Director of Engineering at Delivery Hero, and IT Manager at yemeksepeti.com. They have a background in software engineering, system-business analysis, and project management, with a master's degree in Computer Engineering. Mert has also worked as an IT Project Team Lead and has experience in managing mobile teams and global expansions in the online food ordering industry.

// Related Links

Website: https://www.prosus.com/

Website: https://justeattakeaway.com/

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

Join our Slack community [https://go.mlops.community/slack]

Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

Sign up for the next meetup: [https://go.mlops.community/register]

MLOps Swag/Merch: [https://shop.mlops.community/]

MLOps GPU Guide: https://go.mlops.community/gpuguide

Connect with Demetrios on LinkedIn: /dpbrinkm

Connect with Euro on LinkedIn: /eurobeinat/

Connect with Mert on LinkedIn: /mertoztekin/

Timestamps:

[00:00] AI Transformation Challenges

[00:29] AI Productivity

[04:30] Developer Tool Freedom

[09:40] AI Alignment Bottleneck

[22:17] Exploring Agent Potential

[25:59] Governance of AI Agents

[33:24] Shadow AI Governance

[40:57] AI Budgeting for Growth

[46:27] MLOps GPU Guide announcement!

Zengyi Qin is the Founder of the OpenAGI Foundation, working on computer-use models and open, agent-centric AI infrastructure.

Computers that Think and Take Actions for You, Zengy Qin // MLOps Podcast #355

Join the Community:

https://go.mlops.community/YTJoinIn

Get the newsletter: https://go.mlops.community/YTNewsletter

MLOps Merch:

https://shop.mlops.community/

// Abstract

What if the computer itself can think and take actions for you? You just give it a goal, and it performs every click, type, drag, and gets work done across the desktop and web. In this talk, Zengyi reveals the breakthrough technology that his company OpenAGI is developing: AI that can use computers like humans do. He talks about how his team developed the model, why it outperforms similar models from OpenAI and Google, and its wide use cases across different domains.

// Related Links

Website: https://www.qinzy.tech/

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

Join our Slack community [https://go.mlops.community/slack]Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

Sign up for the next meetup: [https://go.mlops.community/register]

MLOps Swag/Merch: [https://shop.mlops.community/]

Connect with Demetrios on LinkedIn: /dpbrinkm

Connect with Zengyi on LinkedIn: /qinzy/

Timestamps:

[00:00] AI and Human Interaction

[00:30] Zengyi's story

[08:19] Why Expensive Models Lost

[06:30] Bigger Models Are Lazy

[10:24] Training Computer-Use vs LLMs

[13:53] World Models and Sandboxes

[19:42] Dealing with Non-Stationary States

[23:56] Training with Software

[26:44] Sandbox Training Process

[41:33] Infrastructure for Computer Models

[44:36] Wrap up

Varant Zanoyan is the Co-founder & CEO at Zipline AI, working on building a next-generation AI/ML infrastructure platform that streamlines data pipelines, model deployment, observability, and governance to accelerate enterprise AI development.

Nikhil Simha Raprolu is the Co-founder & CTO at Zipline AI, focused on architecting and scaling the company’s AI data platform — extending the open-source Chronon engine into a developer-friendly system that simplifies building and operating production AI applications.

Real-time features, AI search, Agentic similarities, Varant Zanoyan & Nikhil Simha Raprolu // MLOps Podcast #354

Join the Community:

https://go.mlops.community/YTJoinIn

Get the newsletter: https://go.mlops.community/YTNewsletter

MLOps Swag/Merch:

[https://shop.mlops.community/]

And huge thanks to Chroma for hosting us in their recording studio

// Abstract

Feature stores might be the wrong abstraction. Varant Zanoyan and Nikhil Simha Raprolu explain why Cronon ditched “store-first” thinking and focused on compute, orchestration, and real-time correctness—born at Airbnb, battle-tested with Stripe. If embeddings, agents, and real-time ML feel painful, this episode explains why.

// Related Links

Website: https://zipline.ai/

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

Join our Slack community [https://go.mlops.community/slack]

Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

Sign up for the next meetup: [https://go.mlops.community/register]

MLOps Swag/Merch: [https://shop.mlops.community/]

Connect with Demetrios on LinkedIn: /dpbrinkm

Connect with Varant on LinkedIn: /vzanoyan/

Connect with Nikhil on LinkedIn: /nikhilsimha/

Timestamps:

[00:00] Feature Platform Insights

[02:00] Zipline and Feature Stores

[05:19] Cronon and Zipline Origins

[10:49] Feast and Feather Comparison

[13:27] Open source challenges

[20:52] Zipline and Iceberg Integration

[23:54] Airbnb Agent Systems

[28:16] Features vs Embeddings

[29:07] Wrap up

Alex Salazar is the CEO and Co-Founder of Arcade.dev, working on secure AI agents and real-world automation integrations.

Chiara Caratelli is a Data Scientist at Prosus Group, working on AI agents, web automation, and evaluation of robust multimodal models.

Join the Community:

https://go.mlops.community/YTJoinIn

Get the newsletter: https://go.mlops.community/YTNewsletter

MLOps GPU Guide: ⁠

https://go.mlops.community/gpuguide

// Abstract

Agents sound smart until millions of users show up. A real talk on tools, UX, and why autonomy is overrated.

// Bio

Chiara Caratelli

Chiara is a Data Scientist at Prosus, where she develops AI-driven solutions with a focus on AI agents, multimodal models, and new user experiences. With a PhD in Computational Science and a background in machine learning engineering and data science, she has worked on deploying AI-powered applications at scale, collaborating with Prosus portfolio companies to drive real-world impact.

Beyond her work at Prosus, she enjoys experimenting with generative AI and art. She is also an avid climber and book reader, always eager to explore new ideas and share knowledge with the AI and ML community.

Alex Salazar

Alex is the CEO and co-founder of Arcade.dev, the unified agent action platform that makes AI agents production-ready. Previously, Salazar co-founded Stormpath, the first authentication API for developers, which was acquired by Okta. At Okta, he led developer products, accounting for 25% of total bookings, and launched a new auth-centric proxy server product that reached $9M in revenue within a year. He also managed Okta's network of over 7,000 auth integrations. Alex holds a computer science degree from Georgia Tech and an MBA from Stanford University.

// Related Links

Website: https://www.prosus.com/

Website: https://www.arcade.dev/

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

Join our Slack community [https://go.mlops.community/slack]

Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

Sign up for the next meetup: [https://go.mlops.community/register]

MLOps Swag/Merch: [https://shop.mlops.community/]

Connect with Demetrios on LinkedIn: /dpbrinkm

Connect with Alex on LinkedIn: /alexsalazar/

Connect with Chiara on LinkedIn: /chiara-caratelli/

Timestamps:

[00:00] Intro

[00:15] Insights from iFood

[06:22] API vs agent intention

[09:45] Tool definition clarity

[15:37] Preemptive context loading

[27:50] Contextualizing agent data

[33:27] Prompt bloat in payments

[41:33] Agent building evolution

[50:09] Agent program scalability

[55:29] Why multi-agent is a dead end

[56:17] Wrap up

Jonathan Wall is the CEO at Runloop.ai, working on enterprise-grade infrastructure and execution environments for AI coding agents.

The Future of AI Agents is Sandboxed // MLOps Podcast #353 with Jonathan Wall, CEO at Runloop.ai.

Join the Community:

https://go.mlops.community/YTJoinIn

Get the newsletter: https://go.mlops.community/YTNewsletter

Shoutout to  @runloop-ai  for powering this MLOps Podcast episode.

// Abstract

Everyone’s arguing about agents. Jonathan Wall says the real fight is about sandboxes, isolation, and why most “agent platforms” are doing it wrong.

// Bio

Jon was the techlead of Google File System, a founding engineer at Google Wallet, and then the founder of Inde, which was acquired by Stripe. He is building Runloop.ai to bridge the production gap for AI Agents by building a one-stop sandbox infrastructure for building, deploying, and refining agents.

// Related Links

Website: runloop.ai

Blogs and content at https://www.runloop.ai/

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

Join our Slack community [https://go.mlops.community/slack]

Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

Sign up for the next meetup: [https://go.mlops.community/register]

MLOps Swag/Merch: [https://shop.mlops.community/]

Connect with Demetrios on LinkedIn: /dpbrinkm

Connect with Jon on LinkedIn: /jonathantwall/

Timestamps:

[00:00] GitHubification of workflows

[00:29] Sandbox definitions explained

[04:47] Agent setup explanation

[08:03] Sandbox vs API agent

[13:51] Resource usage in sandbox

[22:50] Agent evaluation setup

[28:08] Failure cases value

[31:06] Sandbox isolation vs multi-tenancy

[36:14] Frameworks vs Harnesses

[39:02] Langraph vs Harness comparison

[43:22] Agent flexibility and verification

[52:51] Training data focus

[57:10] Wrap up