Activity
7K followers
-
Tushar Katarki shared thisI recently sat down with Mike Vizard to discuss one of the biggest challenges enterprises face with AI: how to scale from promising pilots to reliable production deployments. The conversation explores the tension between unlocking AI-driven innovation while managing institutional risk, governance, reliability, and operational complexity. That’s where many organizations are today. I hope you’ll check it out and let me know what resonates with you. P.S. Please excuse the casual “office” setting—I only realized at the last minute that this was going to be published as a video. 😜 https://lnkd.in/gNqVdfs2
-
Tushar Katarki posted thisMeta just dropped Muse Glimmer 30B — its first open-weights model since Llama 4 — and Red Hat AI Inference had Day-0 support ready the same day. Our engineering team worked upstream with Meta and the vLLM community ahead of launch so Muse Glimmer runs natively on vLLM from release day, plus we shipped an FP8-block quantized variant (built with LLM Compressor) at roughly half the memory and disk footprint with comparable accuracy - no compromises needed there when it comes to accuracy. Try it today: HuggingFace: RedHatAI/Muse-Glimmer-30B-FP8-block Catalog link: rhaii-preview/vllm-cuda-rhel9 Open models move fast. vLLM and Red Hat AI Inference are built to keep up. #RedHatAI #vLLM #OpenSource #AIInference
-
Tushar Katarki shared thisWhat does an AI-native product team team actually look like? Not fewer people. Not a chatbot bolted onto old workflows. A rebuilt loop — the same signal-to-decision-to-build-to-ship cycle every product org runs, redesigned stage by stage so agents do meaningful work at every step and humans supervise and decide. I've spent the last year living this from the inside, leading Product for Red Hat AI, as we rebuilt our own development lifecycle around this idea. A few things surprised me: individual productivity gains don't automatically become team or org gains. The messiest phase of adoption — dozens of teams experimenting in parallel — turned out to be necessary, not a mistake. How do you scale that across individuals and cross-functional teams. And the next problem is already visible: as agents start handing work to other agents instead of just to humans, standardization gets more urgent, not less. I'll walk through the actual rebuild, stage by stage, on the main stage at CPO Boston, August 20. If you're rethinking how your product org needs to work in the age of AI agents, I'd love to see you there. https://lnkd.in/gT3qCnuB #ProductManagement #AI #AgenticAI #RedHat #CPOBostonAgenda | Chief Product Officer Summit Boston | Chief Product Officer Summit | BostonAgenda | Chief Product Officer Summit Boston | Chief Product Officer Summit | Boston
-
Tushar Katarki shared thisReally proud to see Red Hat and IBM take a major step forward with Project Lightwell. Open source powers modern enterprise infrastructure and AI, but securing the software supply chain at global scale is becoming one of the defining challenges of this era. Project Lightwell aims to bring a new model for open source security: • AI-assisted identification and remediation of vulnerabilities • A trusted “clearinghouse” for coordinated fixes and validation • Secure software supply chain workflows across development through production • Broader protection for OSS components, libraries and AI frameworks used by enterprises worldwide This is exactly the kind of industry leadership and open collaboration that makes enterprise open source so powerful. Heartfelt congratulations to the leadership, engineers, security teams and open source communities helping make this happen. Excited to see where this goes next. #opensource #security #ai #redhat #ibm #supplychainsecurity https://lnkd.in/e47NENdRIBM and Red Hat Commit $5 Billion to Redefine the Future of Open Source in the AI EraIBM and Red Hat Commit $5 Billion to Redefine the Future of Open Source in the AI Era
-
Tushar Katarki shared thisThis extends the idea of “hybrid AI” to the next level. On one end of the spectrum, organizations can build fully self-hosted agentic systems — model to runtime to tool execution — entirely on their own infrastructure. That path is compelling, and over time many enterprises will move in that direction for control, governance, and sovereignty. But there’s another very pragmatic model emerging as well: Cloud-managed agents (for example from providers like Anthropic) combined with actual tool execution, data access, and policy enforcement running securely inside customer environments. That’s where technologies like OpenShell, Kata Containers, Kagenti, and platforms like Red Hat AI become especially interesting — enabling enterprises to combine the agility of managed frontier models with on-premises security boundaries, guardrails, observability, and governance. The future of “hybrid” may not just be where models run — but where reasoning, actions, tools, and trust boundaries are distributed.Tushar Katarki shared this“The best AI reasoning in the world should not require you to give up control of your execution environment. Outsource the thinking. Keep the doing.” Bringing agentic code execution to the hybrid cloud with Anthropic self-hosted sandboxes in OpenShell with Red Hat AI. Derek Carr Mrunal Patel Adel ZaaloukBringing Claude self-hosted sandboxes to OpenShell on Red Hat AIBringing Claude self-hosted sandboxes to OpenShell on Red Hat AI
-
Tushar Katarki shared thisFantastic opportunities to work on cutting edge technologies that are shaping red hat ai platform for our customers and this is an amazing team too!Tushar Katarki shared thisWe're hiring in Boston (hybrid) at two levels: - Principal Software Engineer --- 8+ years, technical leadership, you've shipped complex systems and can drive architecture decisions across squads. Multiple openings. - Software Engineer --- Earlier career, strong ML fundamentals, you learn fast and want to grow in a research-adjacent environment. Our team works across multiple areas of LLMs and agents, including post-training, synthetic data, inference-time scaling, reinforcement learning, and agent harnesses. We invent new technologies, turn them into open-source software, and ship them on the Red Hat AI platform. We also publish papers and make everything open source. This year alone, our ~10-person team published 4 ICML papers, 1 ICLR paper, 1 ACL paper, and 2 AAAI Oral papers. We're looking for people who: - Write solid, production-quality code and understand what's happening under the hood - Understand distributed systems and can build on Kubernetes at scale - Care about open source and want to grow with us! More about us: https://ai-innovation.team Apply: - Principal SWE: https://lnkd.in/eqpH3qaM - Software Engineer: https://lnkd.in/ecjBtVBi Or DM me directly --- always happy to talk about what we're building. #RedHat #AIEngineering #OpenSource #MachineLearning #Hiring #LLM #Boston
-
Tushar Katarki shared thisAs I head to Red Hat Summit Atlanta starting May 11, one thing I keep thinking about is how quickly the worlds of AI agents and cybersecurity are converging. A lot of the current discussion around “agent security” sounds new: sandboxing, least privilege, runtime governance, zero trust, policy enforcement, behavioral monitoring. But in many ways, cybersecurity has been solving versions of this problem for years. Modern infrastructure already assumes that software can become unpredictable — whether through exploits, supply chain compromise, stolen credentials, or malicious automation. That’s why we built layers like containers, Kubernetes isolation, SELinux, RBAC, service mesh, workload identity, runtime observability, and zero trust architectures. AI agents amplify the same challenge. The difference is that agents are inherently capable systems: they reason, plan, generate code, chain tools together, and take actions dynamically at runtime. Which means the future may not be about “securing AI” separately from “securing applications.” Instead, it may be about continuously governing autonomous actors — whether they are: - AI agents - CI/CD systems - Kubernetes operators - automation pipelines - APIs - or traditional applications themselves Looking forward to digging into this and a lot more at Red Hat Summit. You can find me in these three sessions and on the show floor and in the corridors or maybe at a pub after hours ! - Red Hat AI: What’s new and what’s next, from production inference to autonomous agents - Industrializing enterprise AI: Optimizing rack-scale performance with Red Hat AI Factory with NVIDIA - Scaling LLM inference at a large bank with llm-d and Red Hat OpenShift AI https://lnkd.in/ezMiUimN
-
Tushar Katarki shared thisI was listening to a New York Times Daily episode on my drive to work earlier today … and it genuinely stopped me in my tracks. Not because it was hype. But because it felt real. Before you read any further — if you can, stop here and listen to the episode first. It’s worth it. Two things really stayed with me. First — this isn’t theory anymore. The piece was based on conversations with 75+ software developers across startups and large companies. And what they described didn’t sound like the future. It sounded like… right now: - Writing far less code (sometimes almost none) - Letting AI handle large parts of day-to-day work - Moving dramatically faster in small teams - Spending more time thinking, iterating, and guiding vs. typing What surprised me most wasn’t the change itself. It was the tone. What surprised me most wasn’t just the shift — it was how quickly people adapted. There was one developer (Manu) who started out pretty skeptical. Worried about hallucinations, bad code, all the usual concerns. And then… he leaned all the way in. He started treating the AI almost like a team — even writing very stern, almost comical system prompts: - “You MUST run these tests.” - “Failure to do this is unacceptable and embarrassing.” And somehow… it worked. That part really stuck with me. Not because it’s funny (it is), but because it shows how people are figuring out new ways of working in real time. Most of them didn’t sound anxious. They sounded… energized. That feeling of building something is still there — just happening differently. Second — when The New York Times is telling this story, it’s no longer niche. This isn’t a tech bubble conversation anymore. It’s a mainstream publication documenting how work is already changing — through real people, real workflows. I’ve seen a few platform shifts up close over the years. There’s always a phase where it sounds big… and then a moment where it quietly becomes normal. This felt like that moment. What stayed with me after that drive: - AI isn’t just making people faster. - It’s changing what it means to do the job. And if that’s happening this deeply in software — one of the most technical fields we have — it’s hard to imagine it stays contained there. I feel good and energized! https://lnkd.in/eUCjWDNS
-
Tushar Katarki reposted thisAwesome opportunity for a young professional to be exposed to leading edge technologies and a team that will really push you to become your best.Tushar Katarki reposted thisREPOSTING BECAUSE STILL LOOKING I'm looking to hire a recent college grad or an early-career professional to join the Technical Marketing squad. If you like diving into technology but then coming up for air to explain it clearly, this role is for you. You can apply here: https://lnkd.in/eEt3qPUFAssociate Technical Marketing Manager - Application PlatformAssociate Technical Marketing Manager - Application Platform
Tushar Katarki commented on a post
1w
Thanks for moderating and hosting. What an amazing turnout and enjoyed it
Tushar Katarki commented on a post
2w
💯 I mean come on … it’s as if there are no adults when the equities markets sends signals as it has about software. Sure one could argue that fancy software guis may not be needed … but to say that the entire business software stack can be replaced with vibe code is silly
-
Tushar Katarki liked thisTushar Katarki liked this🤖 Enterprise AI is moving from pilot projects into production. On Techstrong.ai Leadership Insights, Michael Vizard speaks with Tushar Katarki, Head of Product for Red Hat AI Platforms, about what it takes to run AI securely, efficiently and at scale across the enterprise IT stack. 🌐 Katarki explains why open source models and platforms are giving organizations more control over cost, data, infrastructure and sovereignty as geopolitical shifts reshape technology strategy. ⚙️ The conversation also explores Red Hat’s investment in vLLM as an open source inference engine, along with llm-d for distributed inferencing across different models, workloads and AI accelerators. 🔐 As AI agents move beyond chatbots into longer-running workflows, enterprises need AI gateways with token tracking, rate limits, quota management, chargeback, guardrails and tool-calling controls. 🎯 The takeaway: AI is becoming a new class of enterprise workload. IT teams need to operate like model service providers and agent service providers while keeping governance, cost and security under control. 🎥 Watch Mike Vizard’s full interview with Tushar Katarki of Red Hat here: https://lnkd.in/esiJHw6W #AI #EnterpriseAI #RedHat #OpenSource #DataSovereignty #AIInfrastructure #Governance #TechstrongTV
-
Tushar Katarki reacted on thisTushar Katarki reacted on thisAI is moving to the factory floor. But scaling it? That’s where things usually stall. 🛑 Deploying AI at the industrial edge comes with unique challenges—from harsh environments to massive data silos. You can't just copy-paste cloud solutions and hope for the best. That’s why Red Hat and Intel teamed up to build a repeatable, blueprinted path to production. By combining Red Hat OpenShift with Intel’s edge-optimized hardware and OpenVINO toolkit, we're helping enterprises: 🛠️ Simplify deployments across diverse edge environments. 📈 Scale reliably without rebuilding infrastructure. ⚡ Turn raw edge data into real-time operational efficiency. Ready to take your edge AI from a cool pilot to a repeatable powerhouse? 👉 Read the full blog here: https://lnkd.in/gNYuAHec #IndustrialEdge #EdgeAI #RedHat #Intel #SmartManufacturing Vinodh Raghunathan, PhD Cole Wangsness Joshua David Kelly Switt Wei Yeang (WY) Toh Edel Curley Abhijit SinhaAI in production at the industrial edge: A repeatable path with Red Hat and IntelAI in production at the industrial edge: A repeatable path with Red Hat and Intel
-
Tushar Katarki liked thisIf you’re in town on 24 October 2026, don’t miss this talk by Alex Lee on Red Hat AI Inference on AWS Inferentia. A great opportunity to explore the intersection of vLLM, Red Hat AI, AWS Inferentia, and production scale generative AI, and learn how modern inference stacks can deliver better performance and efficiency. Steve Shirkey Sean W.M. Chan Albert Law Peter Man 文志鋒 Tushar Katarki #redhat #aws #vllm #redhataiTushar Katarki liked thisWhat will happen for Red Hat vLLM in Amazon Web Services (AWS) Neuron with AWS Inferentia? #redhat Senior SA Alex Lee will share it in #awscommunityday #hongkong Don't miss the session and join now: https://lnkd.in/gcxx2-XN #thankyou Li Ming Tsai to lineup! Hong Kong Institute of Information Technology (HKIIT)
-
Tushar Katarki liked thisTushar Katarki liked thisYou might think VMs are just for legacy infrastructure. Surprise: they’ve had a major glow-up. On Compiler, Red Hat’s Maria Bracho explains how virtualization doesn’t just keep the lights on for older tech–and how it has become a key part of AI infrastructure: https://red.ht/4cGPRDu.
-
Tushar Katarki reacted on thisTushar Katarki reacted on thisI appreciate being included in TIME's annual list of the 100 most influential people in AI globally. Let's keep the AI for Science and the Genesis Mission momentum going! https://lnkd.in/gegeRNbH
-
Tushar Katarki reacted on thisTushar Katarki reacted on thisThe call for collective cyber defense (https://lnkd.in/gd232873) published today gets one thing exactly right: security in the AI era cannot be solved in isolation. Red Hat signed because collaboration isn’t a new security strategy for us. It’s foundational to how open source works. The call to action is important. What comes next is even more important. We need shared, collaborative ways to learn from security incidents involving AI models and agents, so the lessons from one incident can strengthen defenses across the ecosystem. And we need to work together to build the open tools enterprises will need to protect themselves as AI reshapes cybersecurity. As AI accelerates vulnerability discovery and remediation, that model matters more, not less. Collective defense requires trusted collaboration across researchers, maintainers, vendors and enterprises, along with the engineering expertise to turn faster discovery into fixes organizations can confidently deploy. We’re already putting those principles into practice. With Lightwell, we’re applying AI and engineering expertise to vulnerability remediation while maintaining our commitment to working upstream. And with efforts like OpenShell, we’re working to advance security for the emerging world of AI agents. These are different challenges, but the principle is the same: open collaboration can help the industry learn faster, build better defenses, and strengthen security for everyone.
-
Tushar Katarki liked thisTushar Katarki liked thisWe are honored to welcome Vijay Vusirikala, Distinguished Lead, AI Systems & Networks at Arista, as a featured panelist at the MantisGrid AI Hackathon & Summit 2026 on September 17 in Palo Alto. As AI clusters scale to thousands of GPUs, networking is becoming one of the foundational elements of AI infrastructure—critical to performance, reliability, and the ability to operate AI systems at massive scale. Vijay brings a unique perspective at the intersection of AI systems, hyperscale infrastructure, and networking. At Arista, Vijay focuses on networking solutions for hyperscalers and large AI clusters. Previously, he led global network engineering at an AI infrastructure startup and spent 14+ years at Google in senior technical and organizational leadership roles, helping develop the network stack behind one of the world’s largest global backbone networks. With 17+ patents and decades of experience building and scaling critical network infrastructure, Vijay will bring a deeply technical and real-world perspective to our “The Future of AI Infrastructure” panel—exploring how networking architectures must evolve as AI clusters and workloads continue to grow. 📅 September 17, 2026 | 8:00 AM–8:00 PM PT 📍 Mitchell Park Community Center, Palo Alto 🎟️ $10,000 Hackathon Prize Pool 🔗 Register: https://luma.com/d6g5mb9o — Limited seats are now available. We are excited to have Vijay join an outstanding group of AI infrastructure leaders, builders, and innovators shaping what comes next. MantisGrid AI, Inc Vijay Vusirikala Randolph Chung #MantisGridAI #AIInfrastructure #AINetworking #Arista #ArtificialIntelligence #AI #Hackathon #PaloAlto
-
Tushar Katarki liked thisTushar Katarki liked thisI am excited to share that I recently wrapped up my final week as a Biomedical Engineering Intern at Upsilon Health. It was an honor to work at such an innovative women’s health company whose mission is to make contraceptive health more safe and accessible via a novel copper and hormone free IUD. Over the course of thirteen weeks, I gained incredible hands-on engineering experience while learning the ins and outs of startups. This summer, my key projects included: 1. Designing a fixture for mechanical testing of the IUD using Computer Aided Design (CAD) 2. Writing verification and validation testing procedures 3. Meeting with suppliers and gathering quotes to guide business decisions 4. Supporting budget planning for a grant submission. Thank you so much to Richard Briganti for mentoring me in engineering design, manufacturing, and product development, and Upsilon’s incredible female founder, Antonella Sturniolo-DePue, MPH. Thank you to the entire Upsilon team for an amazing summer full of learning and innovation!
Experience & Education
Languages
-
Hindi
-
Recommendations received
6 people have recommended Tushar
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content