Activity
29K followers
-
Hart Rossman reposted thisHart Rossman reposted thisOne of the most special days of the year has arrived: Happy First Day of Classes, Terps! 🐢
-
Hart Rossman reposted thisHart Rossman reposted this𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗮𝘀𝘀𝗲𝘀𝘀𝗺𝗲𝗻𝘁𝘀 𝘄𝗶𝘁𝗵 𝗳𝗿𝗼𝗻𝘁𝗶𝗲𝗿 𝗔𝗜 𝗮𝗿𝗲 𝗲𝘅𝗽𝗹𝗼𝗿𝗮𝘁𝗼𝗿𝘆, 𝗻𝗼𝘁 𝗱𝗲𝘁𝗲𝗿𝗺𝗶𝗻𝗶𝘀𝘁𝗶𝗰. 𝗧𝗵𝗮𝘁 𝗵𝗮𝘀 𝗺𝗮𝗷𝗼𝗿 𝗶𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 𝗳𝗼𝗿 𝗯𝗼𝘁𝗵 𝗰𝗼𝘃𝗲𝗿𝗮𝗴𝗲 𝗮𝗻𝗱 𝗰𝗼𝘀𝘁. In the latest installment of our 𝘍𝘳𝘰𝘯𝘵𝘪𝘦𝘳 𝘈𝘐 𝘍𝘪𝘦𝘭𝘥 𝘕𝘰𝘵𝘦𝘴 series, Sean McRae shows how little prompting is needed for newer models to continuously explore new paths through the same codebase. 𝗧𝗵𝗲 𝗰𝗵𝗮𝗿𝘁 𝗯𝗲𝗹𝗼𝘄 𝘀𝗵𝗼𝘄𝘀 𝘁𝗵𝗲 𝟭𝟬𝘁𝗵 𝗿𝘂𝗻 𝘀𝘁𝗶𝗹𝗹 𝘂𝗻𝗰𝗼𝘃𝗲𝗿𝗲𝗱 𝟭𝟵 𝗳𝗶𝗻𝗱𝗶𝗻𝗴𝘀 𝘁𝗵𝗮𝘁 𝗻𝗼𝗻𝗲 𝗼𝗳 𝘁𝗵𝗲 𝗽𝗿𝗲𝘃𝗶𝗼𝘂𝘀 𝗻𝗶𝗻𝗲 𝗿𝘂𝗻𝘀 𝗵𝗮𝗱 𝗶𝗱𝗲𝗻𝘁𝗶𝗳𝗶𝗲𝗱. If you thought you had reached saturation at run six, you would have been wrong. 𝗧𝗵𝗶𝘀 𝗰𝗿𝗲𝗮𝘁𝗲𝘀 𝗮 𝗽𝗿𝗮𝗰𝘁𝗶𝗰𝗮𝗹 𝗰𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲 𝗳𝗼𝗿 𝗮𝗻𝘆𝗼𝗻𝗲 𝘂𝘀𝗶𝗻𝗴 𝗔𝗜-𝗮𝘀𝘀𝗶𝘀𝘁𝗲𝗱 𝘀𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗮𝘀𝘀𝗲𝘀𝘀𝗺𝗲𝗻𝘁𝘀: • How many runs are enough? • How do we know when we've achieved reasonable coverage? • If additional runs continue producing new findings, what are the economics of stopping too early versus continuing to search? We'd love to hear how others are thinking about this problem. 𝘓𝘪𝘯𝘬 𝘵𝘰 𝘚𝘦𝘢𝘯'𝘴 𝘧𝘪𝘦𝘭𝘥 𝘯𝘰𝘵𝘦 𝘪𝘯 𝘵𝘩𝘦 𝘤𝘰𝘮𝘮𝘦𝘯𝘵𝘴. #AI #Cybersecurity #ModelEvaluation #FrontierAI #Glasswing #FrontierAIFieldNotes
-
Hart Rossman reposted thisHart Rossman reposted thisWe are excited to announce that Hart Rossman, VP AWS Security, is joining the 2026 CRO Risk Summit as a guest speaker. His presentation, "Cyber, Quantum and Frontier Models: How the World is Changing" will help participants navigate the ever-evolving risks leaders face in business today. The Summit is an exclusive event for senior risk leaders across industries. If you are a CRO or C-suite executive interested in participating, contact Mary Bittle (Teer) Koenick for registration information. University of Maryland - Robert H. Smith School of Business
-
Hart Rossman reposted thisHart Rossman reposted thisHow do you screen 27 prompt changes in a week and still know which result belongs to which change? I turned the Kiro team's detailed write-up into this one-page diagram. They write the candidate changes and use the results to decide what ships. An automated system runs the evaluation. The LLM judge scores internal developer conversations across 15 behavioral dimensions using explicit evidence, and then eligible developers stay in stable cohorts so each candidate can be compared with the control. The same rubric scores every group, and a combined cohort checks interactions. Small cohorts help locate regressions, but the team needs larger follow-ups before drawing broader conclusions. When the model changes, the team runs the checks again. That's why the 32% and 4% results are separate: they came from different model-and-prompt baselines. Full write-up from the eng + research teams: https://lnkd.in/gFh2iZua 👩🏻💻 Follow me ( Brooke Jamieson ) for the latest AI + AWS updates 📍 Save + Share 🏷️ AWS Developers #Kiro #DeveloperTools #AICoding #AgenticAI 🐈 This is a clear and conspicuous disclaimer that I am an AWS employee and all opinions are my own 🐈
-
Hart Rossman reposted thisHart Rossman reposted thisYesterday at the AWS Startup Partner Summit in San Francisco, we launched Startups You Should Know, a new view in the Solution Matching Engine that gives AWS sellers a dedicated way to discover startup partners -- even before they have a long track record on AWS. Co-sell is an AWS superpower. Just ask the growing number of partners who've transacted over $1B on AWS Marketplace, or startups that have closed company-changing enterprise deals alongside our sales teams. Solution Matching Engine uses AI to match the right partner solutions to customer opportunities, so sellers find the right partner faster. Learn more: https://lnkd.in/eDuV32hP
-
Hart Rossman reposted thisHart Rossman reposted thisVerus verifies complex projects with thousands of lines of code and proof in the time prior tools took to verify individual functions. That speed enables an interactive development loop and lets AI agents iterate on proofs faster. Amazon has used this open-source, automated program verifier for Rust to prove correctness of key primitives in the Nitro Isolation Engine and critical internal infrastructure.
-
Hart Rossman reposted thisHart Rossman reposted thisProud to see the University of Maryland recognized among the nation’s top public universities! In the inaugural edition of TIME and Statista’s America’s Best Colleges 2026–2027, UMD ranked No. 24 among “public” institutions nationwide (No. 63 overall). Maryland also placed among the top 20 public universities for student outcomes (a measure based on graduation rates, graduate earnings and return on investment). This recognition reflects what Terps already know: a Maryland education delivers tremendous value while preparing students to compete, lead and thrive after graduation. Proud to be a Terp!
-
Hart Rossman reposted thisHart Rossman reposted thisOnce again we expand our open standard multicloud interconnection solution by adding Azure! Been great working with them to get this ready for customers! If you have multicloud workloads give it a try as it’s as simple as a few clicks to spin up capacity with 4 regions to start with and many more coming! https://lnkd.in/gEXUxfUvAWS and Microsoft Azure collaborate to expand multicloud networking | Amazon Web ServicesAWS and Microsoft Azure collaborate to expand multicloud networking | Amazon Web Services
-
Hart Rossman reposted thisHart Rossman reposted thisAWS Agent Registry is now generally available. As agents scale from dozens to thousands, you need a single place to discover what exists, govern who can use it, and trace every action, regardless of where those agents run, on any cloud or on-prem. Southwest Airlines went from dozens of agents scattered across teams to a single governed catalog that cut duplicative development effort and let developers find approved capabilities in seconds. PepsiCo is using it to centralize discovery and governance across the enterprise. Syngenta's teams publish once, then reuse what already works instead of rebuilding from scratch. Governance is what gets agents from pilot to production at scale: https://lnkd.in/ebrjZPhC
Hart Rossman commented on a post
9h
Welcome back!!
-
Hart Rossman liked thisHart Rossman liked thisOne of the most special days of the year has arrived: Happy First Day of Classes, Terps! 🐢
-
Hart Rossman liked thisHart Rossman liked thisExcited to share that I've returned to Amazon to lead research for Principal Engineering as part of Executive Talent Acquisition. It's been so great to reconnect with former colleagues and meet many new ones! We are hiring: https://lnkd.in/gnr_k4iS
-
Hart Rossman liked thisHart Rossman liked this𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗮𝘀𝘀𝗲𝘀𝘀𝗺𝗲𝗻𝘁𝘀 𝘄𝗶𝘁𝗵 𝗳𝗿𝗼𝗻𝘁𝗶𝗲𝗿 𝗔𝗜 𝗮𝗿𝗲 𝗲𝘅𝗽𝗹𝗼𝗿𝗮𝘁𝗼𝗿𝘆, 𝗻𝗼𝘁 𝗱𝗲𝘁𝗲𝗿𝗺𝗶𝗻𝗶𝘀𝘁𝗶𝗰. 𝗧𝗵𝗮𝘁 𝗵𝗮𝘀 𝗺𝗮𝗷𝗼𝗿 𝗶𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 𝗳𝗼𝗿 𝗯𝗼𝘁𝗵 𝗰𝗼𝘃𝗲𝗿𝗮𝗴𝗲 𝗮𝗻𝗱 𝗰𝗼𝘀𝘁. In the latest installment of our 𝘍𝘳𝘰𝘯𝘵𝘪𝘦𝘳 𝘈𝘐 𝘍𝘪𝘦𝘭𝘥 𝘕𝘰𝘵𝘦𝘴 series, Sean McRae shows how little prompting is needed for newer models to continuously explore new paths through the same codebase. 𝗧𝗵𝗲 𝗰𝗵𝗮𝗿𝘁 𝗯𝗲𝗹𝗼𝘄 𝘀𝗵𝗼𝘄𝘀 𝘁𝗵𝗲 𝟭𝟬𝘁𝗵 𝗿𝘂𝗻 𝘀𝘁𝗶𝗹𝗹 𝘂𝗻𝗰𝗼𝘃𝗲𝗿𝗲𝗱 𝟭𝟵 𝗳𝗶𝗻𝗱𝗶𝗻𝗴𝘀 𝘁𝗵𝗮𝘁 𝗻𝗼𝗻𝗲 𝗼𝗳 𝘁𝗵𝗲 𝗽𝗿𝗲𝘃𝗶𝗼𝘂𝘀 𝗻𝗶𝗻𝗲 𝗿𝘂𝗻𝘀 𝗵𝗮𝗱 𝗶𝗱𝗲𝗻𝘁𝗶𝗳𝗶𝗲𝗱. If you thought you had reached saturation at run six, you would have been wrong. 𝗧𝗵𝗶𝘀 𝗰𝗿𝗲𝗮𝘁𝗲𝘀 𝗮 𝗽𝗿𝗮𝗰𝘁𝗶𝗰𝗮𝗹 𝗰𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲 𝗳𝗼𝗿 𝗮𝗻𝘆𝗼𝗻𝗲 𝘂𝘀𝗶𝗻𝗴 𝗔𝗜-𝗮𝘀𝘀𝗶𝘀𝘁𝗲𝗱 𝘀𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗮𝘀𝘀𝗲𝘀𝘀𝗺𝗲𝗻𝘁𝘀: • How many runs are enough? • How do we know when we've achieved reasonable coverage? • If additional runs continue producing new findings, what are the economics of stopping too early versus continuing to search? We'd love to hear how others are thinking about this problem. 𝘓𝘪𝘯𝘬 𝘵𝘰 𝘚𝘦𝘢𝘯'𝘴 𝘧𝘪𝘦𝘭𝘥 𝘯𝘰𝘵𝘦 𝘪𝘯 𝘵𝘩𝘦 𝘤𝘰𝘮𝘮𝘦𝘯𝘵𝘴. #AI #Cybersecurity #ModelEvaluation #FrontierAI #Glasswing #FrontierAIFieldNotes
-
Hart Rossman reacted on thisHart Rossman reacted on thisIt is hard to believe yesterday was my last day at AWS, a lot changed in the 14 years since I joined AWS (getting married, having 3 kids, moving to Europe). It is sad to say goodbye but excited to have the next chapter back with the family in the USA (San Francisco). A huge thank you to all my teams, leaders, customers, partners over the years - way too many to tag. Excited for the new adventure to announce next week. Jon Jones Darren Mowry Kamini Aisola Paul Soligon Ed Porter David Cristini
-
Hart Rossman liked thisHart Rossman liked thisYesterday we kicked off the AWS Startup Partner Summit, and there was a consistent theme from several of the discussions: startups are thinking differently about what partnership with AWS can mean for their business. I sat down with Scott Stephenson, CEO and Co-founder of Deepgram , to talk about the company’s journey with AWS. What started with infrastructure and engineers working together to solve hard technical problems has grown into a much broader partnership as Deepgram has scaled. That evolution is something we’re seeing across the startup community. Founders want more than great infrastructure. They want partners who can help them drive outcomes with their customers, shorten the path from building to selling, and ultimately grow their business. It’s also an area where we continue to lean into. AWS Marketplace is increasingly becoming an important route for startups selling to enterprise customers. And through tools like “Startups You Should Know” via our Solutions Matching Engine, we’re working to make it easier for AWS customers to discover startups with technology that can solve their most important problems. And that’s the real opportunity — helping startups build on AWS, scale on AWS, and increasingly, grow their business effectively. Thanks to Scott for sharing the Deepgram story, and to Ruba Borno and Matt Yanchyshyn for helping us kick off a great first day. #AWS #Startups #Founders #AWSMarketplace
Experience & Education
Publications
-
X-SCM: The New Science of X-treme Supply Chain Management
Routledge
This book brings together an unprecedented array of contributions from world experts in supply chain volatility management. Their collective writings cover the supply chain in all its forms — physical, financial, cyber, and service — and do so in a way that dials up the science of supply chain management to the next level, to a level capable of taking on the kind of extreme oscillation, change, and even crises that characterize our world today.
Other authorsSee publication
Patents
-
Providing secure software project development environments
Issued US-11838296-B1
-
Cloud computing identity ecosystem
Issued US-11785051-B1
-
Robust data tagging
Issued US-11328073-B1
-
Tokenization federation service
Issued US-11089028-B1
-
Quorum-based access management
Issued US-11019068-B2
-
Security appliance provisioning
Issued US-10484331-B1
-
QUORUM-BASED ACCESS MANAGEMENT
Issued US-20190268342-A1
-
Lineage-based trust for virtual machine images
Issued US-10379894-B1
-
Robust data tagging
Issued US-10296750-B1
-
Quorum-based access management
Issued US-10291622-B1
-
Security threat based auto scaling
Issued US-10050999-B1
-
Supplying regulatory compliance evidence for a service provider environment
Issued US-9967285-B1
-
Security recommendation engine
Issued US-9712535-B1
-
Security recommendation engine
Issued US-9386033-B1
Projects
-
SolarPunk Ham Radio
A companion app for the ham radio enthusiast. Built exclusively for the iPhone.
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content