About
Activity
6K followers
-
AJ Kneip reposted thisAJ Kneip reposted thisThere is no safe path to modifying model weights mid-inference. That single constraint splits "continuously learning" into two products sold under one name. The common one is a release schedule. Fine-tuning embeds domain knowledge in the base model weights, so updating that knowledge means going offline, rebuilding from a new checkpoint, and redeploying. Vendors compress the cycle from twelve weeks to six and sell the compression as continuity. The other one requires the adaptation layer to sit outside the base model weights entirely, updating during inference on inference feedback, with the service still running. That is a different architecture rather than a faster version of the same one. Every cycle in the first version is a frozen window. Inside it the system cannot absorb a correction, new data, or a change in domain conditions. When your operational data moves faster than the cycle completes, that becomes a quality ceiling you cannot configure your way out of. The second version is buildable. Separating the adaptation layer is the part we patented. Ghost Weights stores domain adaptation in a reversible parameter overlay held outside the weights. Inline Training updates that overlay mid-inference, per tenant, without disrupting the inference path or the sessions other tenants are running on the same base model. A user corrects an output and the adaptation layer refines on that signal immediately. No six-week cycle, no downtime, no version freeze. Continuous learning has a measurable definition. The time between a user correction and the model behaving differently. Everything else is a release schedule.
-
AJ Kneip reposted thisAJ Kneip reposted thisMost businesses struggle with AI because their AI's starved for context. The institutional knowledge it needs to tap is either woefully incomplete, immediately out of date or downright inaccessible. Our co-founders designed 4Minds to solve all three of those problems *while also* eliminating the data-prep grind and retraining treadmill that stalls out too many AI teams/projects/outcomes. And now (drum roll...), we have third-party validation—from AWS, no less!!—confirming 4Minds can do exactly that. Read what our own AJ Kneip has to say about what this AWS AI Software Competency means and why it matters: https://lnkd.in/gZYkD9rF
-
AJ Kneip shared thisThis one means a lot. ❤️ Six months ago, Michael Kneip Jim Stull convinced me to leave my dream role at Amazon Web Services (AWS) to join an innovative AI lab focused on making business-specific, context-aware, continuously improving intelligence available for every business. Our founders had a simple idea: become the most partner-obsessed AI company on the planet. Today, 4MINDS is officially AWS AI Software Competency Partner. There are a lot of badges you can earn in this industry. This one feels different. It represents months of people showing up, helping each other, taking care of customers, supporting our AWS partners, sweating the details, and believing deeply in what we’re building together. To our team: thank you. To our AWS partners who have believed in us, opened doors, given feedback, and pulled us into the room: thank you. And to our customers who have trusted us with some of their hardest AI problems: thank you!! We’re incredibly proud of this milestone. But mostly, we’re grateful for everyone who helped us get here: Jim Stull Jason Nocco Ryan Trimberger Kubashen Naidoo Ed Porter Tia Wiggins Emily Lewis Kari Seas Robert Ray Kevin Locke Sara Hajjar Eric Yuen Haresh Vaishnav Kaden Freeman Fadi Ghishan Tara Jayne-O'Donnell Michelle Jones Keith Altman Zander Evans Sarah Peters Demaree Barton ☁️ Diego Reyes Jeff Kwarteng Ashley Wax and too many other folks to list!! Six months in. Just getting started. 💙
-
AJ Kneip reposted thisAJ Kneip reposted thisIs GenAI a data-security threat—or a breakthrough business accelerator? 👀 Join us at the CDO Magazine Chicago Leadership Summit for a conversation on turning GenAI into real business impact while navigating security, risk, and adoption. 🎤 Featuring: Ryan Trimberger — 4Minds Dhruv Baronia, CFA — Northern Trust Michael Colella, MS, MA, MS — University of Chicago Sid Raina — Medline Come hear perspectives from leaders across AI, data, analytics, and healthcare on where GenAI is headed—and how organizations can make it work in the real world. 📍 Chicago | September 17, 2026 👉 Save your seat: https://lnkd.in/ex3X-H49 #GenAI #AI #DataLeadership #CDOMagazine #ChicagoLeadershipSummit
-
AJ Kneip reposted thisAJ Kneip reposted thisIf you're wanting to bring AI into your organization, where should you start? A lot of companies pick one department, maybe customer support or marketing, and call it a pilot. Feels like the safe move. But the organizations finding ROI don't silo it. They adopt across the enterprise. When only certain teams get access to AI, you end up with a divisive culture. "So-and-so is more important than I am because they have a model I don't." That's not how you build confidence. Multi-model organizations get this right. When every department has access to models built for how they work, buy-in grows and momentum follows. McKinsey found that organizations seeing real ROI from AI are nearly 3x more likely to have gone enterprise-wide, not just piloted in one or two departments. There's nothing to be afraid of with enterprise-wide AI adoption. This should be exciting. We're looking at a brand new generation of growth. Curious what's stopping your organization from going all in? 📊 𝘚𝘰𝘶𝘳𝘤𝘦: "𝘛𝘩𝘦 𝘚𝘵𝘢𝘵𝘦 𝘰𝘧 𝘈𝘐 𝘪𝘯 2025: 𝘈𝘨𝘦𝘯𝘵𝘴, 𝘐𝘯𝘯𝘰𝘷𝘢𝘵𝘪𝘰𝘯, 𝘢𝘯𝘥 𝘛𝘳𝘢𝘯𝘴𝘧𝘰𝘳𝘮𝘢𝘵𝘪𝘰𝘯"
-
AJ Kneip shared thisGreat week with AWS 💛 From HQ2 Partner Day to an awesome session with the Seattle enterprise team, we talked big customer problems, practical solutions, and what we can build together in AI. Thanks to the amazing and smart people, bold ideas, and all the fun!! Michael Kneip Ed Porter David Jackson Pierre Coffy David Jackson Kaden Freeman Tara Jayne-O'Donnell Haresh Vaishnav Eric Yuen Brian Chu Daniel Huynh Natalie Scarella Tia Wiggins Sara Hajjar ++ SOO many others
-
AJ Kneip reposted thisAJ Kneip reposted thisAttention AWS Sellers 📣 Today 4Minds is joining AWS Arlington Enterprise Partner Day at HQ2. We’re spending the day with the AWS team to collaborate and explore new opportunities for enterprise customers. 4Minds will take part in the Partner Industry Showcase and meet directly with AWS sellers for account mapping, strategy discussions and active opportunities. The day will also include sales enablement sessions and time to connect with AWS leaders and fellow partners. Looking forward to strengthening our partnership with AWS and finding new ways to create value together in H2. Meet up with Ed Porter, Daniel Huynh, Pierre Coffy, and Tia Wiggins if you are there! 📍 AWS HQ2, Arlington, VA, 7th Floor 📅 August 19, 2026
-
AJ Kneip reposted thisAJ Kneip reposted thisYou're paying for your AI twice. The first payment is on the invoice. The second one is your knowledge. A general model is trained on the world. It doesn't know your exceptions or what a correct answer looks like inside your company, so you teach it. And because it runs on the vendor's infrastructure, you use their intelligence while they compound yours. It doesn't feel like disclosure, because it's just you using the product properly. Everyone accepts that a model needs time to get good at your work. Nobody thinks twice about the process. After enough cycles, the value lies with the model provider, not the company that built the expertise. To be clear, I don't think this comes from ill intent. Vendors will tell you they won't train on your data without asking, and I believe them. But that's a policy. And policies change when the business model does. Which is why I think we should view privacy as architecture. It comes down to where the learning happens. Where does the training run? Who owns the model once it knows how you work? If you canceled tomorrow, what leaves with you? We built 4Minds so the answer to all three is you, because when it comes to enterprise AI, you should only have to pay once. 𝘋𝘪𝘢𝘨𝘳𝘢𝘮 𝘣𝘢𝘴𝘦𝘥 𝘰𝘯 𝘚𝘢𝘵𝘺𝘢 𝘕𝘢𝘥𝘦𝘭𝘭𝘢'𝘴 𝘦𝘴𝘴𝘢𝘺 "𝘛𝘩𝘦 𝘙𝘦𝘷𝘦𝘳𝘴𝘦 𝘐𝘯𝘧𝘰𝘳𝘮𝘢𝘵𝘪𝘰𝘯 𝘗𝘢𝘳𝘢𝘥𝘰𝘹"
-
AJ Kneip reposted thisAJ Kneip reposted this60% of companies are getting almost nothing from their AI spend. In contrast, 5% are generating 3.6x higher shareholder returns. Everyone is using the same models. So the model is not the variable. What separates the two groups is that most AI adoption converges. Your competitor buys the same capability, ships it in a quarter of the time, and both of you end up in the same place—with different amounts of money spent. The harder you push down that path, the more you look like everyone else on it. The 5% went in a different direction. They built AI around the one thing a competitor cannot buy: how their business actually works. Knowing a supplier is delayed is information. Understanding how that delay impacts production schedules, customer commitments, and quarterly financials is context. Decisions run on context. Most enterprise AI delivers the first, but not the second—and the gap widens as the business changes. Systems built around context compound. Every disruption resolved and every decision made strengthens what comes next. The advantage doesn't stay fixed at launch; it grows with every decision, every interaction, and every change in the business. Our CEO Ryan Trimberger wrote about how to build that in Harvard Business Review with Haresh Vaishnav and Eric Yuen from Amazon Web Services (AWS): https://lnkd.in/gBZPT5xW
-
AJ Kneip liked thisAJ Kneip liked this🧭 Amazon spent years building chips so it would need Nvidia less. It just ordered two million more of them. AWS and Nvidia said Amazon will deploy an additional 2 million Blackwell Ultra, Rubin and Rubin Ultra GPUs across its infrastructure in 2027 and 2028. Five months ago that commitment was more than 1 million. Nvidia said demand had run past those expectations. Set it against what Amazon has been building in parallel. Trainium is a direct alternative to Nvidia's GPUs. Graviton takes on conventional server silicon from Intel and AMD. Amazon has said its custom chip business passed an annualised revenue pace of $25 billion. Both things are true at once, and that is the interesting part. Building your own silicon has not reduced what you buy from the company you were routing around. It has simply run alongside it. The delivery window is worth noticing too. This capacity lands in 2027 and 2028, not this quarter. It is not an answer to demand today. It is a wager that the demand is still there in two years. So is that genuine conviction about the demand curve, or a hedge nobody in this market can afford not to place? What is your take on this? 🤔💬 Want to stay ahead of where AI is heading? 🔔 Follow AI Insiders for the daily signal on models, agents, and AI products (trusted by 47,636 followers). 📬 Join 14,033 readers of our newsletter and get the AI developments that matter, minus the hype: https://lnkd.in/d_EWhv9k
-
AJ Kneip liked thisAJ Kneip liked thisI’m incredibly excited to share an important evolution at Outreach: Today, our Customer Success Managers officially became GTM AI Advisors! This is more than a title change. It reflects a shift in how we think about Customer Success as AI becomes central to how revenue teams operate. Buying AI and successfully adopting it are two different things, and we believe our customers deserve a strategic partner who owns closing that gap. Our GTM AI Advisors will continue to be the trusted partners our customers already know (supporting platform adoption, EBRs, and renewals as always), while bringing a sharper focus to identifying the right AI-powered workflows, accelerating agent adoption, and connecting AI investment to measurable business outcomes. We're backing this evolution with real capability-building too: every Advisor is completing a formal certification so the title reflects real expertise. I’m so proud of this team and the expertise, curiosity, and commitment they bring to our customers every day. I can’t wait to see the impact they’ll continue to make in this next chapter. Learn more about what this evolution means for our customers: https://lnkd.in/gwbfiDKXMeet the GTM AI Advisor: The Next Evolution of Customer Success | OutreachMeet the GTM AI Advisor: The Next Evolution of Customer Success | Outreach
-
AJ Kneip liked thisAJ Kneip liked thisHad an incredible time trading our dual monitors for panoramic views of Lake Union at the recent team boat cruise in Seattle. Huge thanks Brian Chu and team, for keeping us on an even keel. Mission, a CDW Company is ready to sail into Q4 with some serious momentum!
-
AJ Kneip reacted on thisAJ Kneip reacted on thisHave you ever tried to describe a vision in your head to a designer, a UX team, a software engineer? Every one of those conversations turned into a storyboarding and wireframing cycle, and every cycle came with its own list of problems. Designers and engineers are billable resources, and the business was never eager to spend their time on pre-sales work for a deal that might not close. If the person describing the vision wasn't articulate, the cycles multiplied. If scope shifted mid-build, "oh, I forgot that step" or "we'll need that report too", a change order followed. Costs climbed. QA lost its budget first. Then a project manager was stuck begging for more time, pushing the delivery date, or shelving the project altogether. I lived that for years as a business process automation specialist. I used to write revision cycles directly into SOWs and contracts, just to put a hard limit on it. It never felt fair. I'd bet most of that runaway cost was never about scope. It was about clarity, or the lack of it, before the first wireframe was even drawn. Now I build the storyboard and the wireframe myself, with the customer, in real time. No billable resources spent chasing a deal that hasn't closed yet. The cherry on top is how 4MINDS traces where that data lives, what it's already telling us, and sometimes surfaces a story nobody automated because nobody could see it. #BusinessProcessAutomation #AI #4MINDS #4MINDSAI #DesignInnovation #UI #UX #GenAI #Innovation
-
AJ Kneip reacted on thisAJ Kneip reacted on thisTrying to wrap my head around how active August has been. I keep coming back to gratitude, that I’m so thankful for these experiences and the super cool people I get to spend time with that make these events so impactful. This is an attempt to recap some highlights: 🎙️ Hosted a "State of AI in SaaS" panel + dinner with Ravenna and Aivar. Great conversations, amazing company. 🍽️ Sat down for an engineering exec dinner with the DevZero team. Always energizing to talk shop with people building at that level. 🏝️ Welcomed the Friday Harbor team alongside 4Minds for a session covering agentic systems, evals, and determinism. A full house and great opportunity to slow down and connect. 🛠️ Ran AI Foundry with Aivar, hosting Summation, Strike Graph, and Scala. Awesome to see teams work through real problems together. 🔬 Got to tour the Avalanche Energy research park and learn more about how they’re approaching a new frontier. I also managed to squeeze in a long weekend in Sarasota to be with family and celebrate some recent milestones together. In my moments of free time, I enjoyed the local Seattle greats: Belltown Art Walk, Pioneer Square's Sips in the Square, and a Mariners game. Lastly, one of my favorite colleagues Jon Turdiev and I were recognized for our contributions on a global enablement initiative called Startup AI Interrupt, bringing folks across all of our startup teams to think thru how we tackle some of our org's biggest cross-functional challenges in a world of rapid AI tooling advanvement. A genuine thank you to everyone who humored me and our teams for a workshop, a dinner, a conversation, and/or a too-long doc read. I’m learning to embrace the greater unknown, to celebrate quality connection over nitpicking the marginalia, and to keep challenging myself to dream bigger. Lots more exciting things on the horizon.
-
AJ Kneip reacted on thisAJ Kneip reacted on thisLast week I attended CDO Magazine's Dallas AI in the Enterprise Forum. This was the first event I've attended in my career, and I was honored to hear from so many experienced leaders. The conversations all kept circling back to one theme: governance. Hearing so many different perspectives around the table, and getting to connect with everyone in Dallas, made the day. Our 4Minds CEO moderated a discussion with Thrushna Matharasi, Ranjit Vidhani, and Shuchi Agrawal around a question more enterprises are having to answer: how do you give teams access to powerful AI without giving up control of your data and IP? Being in a room with leaders at this scale this early in my career is something I won't forget. Also, is chocolate milk a regular thing at these events? If so, sign me up for the next one.
Experience & Education
Licenses & Certifications
Recommendations received
2 people have recommended AJ
Join now to viewOther similar profiles
Explore more posts
-
Eddie Reynolds
Thinking of going upmarket? Few know how to rebuild their ops to support it ZoomInfo did. From 50/50 Enterprise/SMB to roughly 70/30. 𝗦𝘂𝗰𝗰𝗲𝘀𝘀 𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗱 𝗼𝗻 𝗿𝗲-𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗚𝗧𝗠 - Enterprise was under-resourced - Outbound was set up for SMB - ICP needed to be redefined - Pipeline needed re-composition - Reps needed different kinds of support 𝗝𝗮𝗺𝗲𝘀 𝗥𝗼𝘁𝗵 𝘄𝗮𝗹𝗸𝗲𝗱 𝘂𝘀 𝘁𝗵𝗿𝗼𝘂𝗴𝗵 𝗶𝘁 - James is the CRO at ZoomInfo - He joined us on the GTM Science podcast - And held nothing back 𝗧𝗵𝗲𝘆 𝗱𝗶𝗱 𝗶𝘁 𝘄𝗶𝘁𝗵 𝗔𝗜, 𝗯𝘂𝗱𝗴𝗲𝘁 𝗿𝗲-𝗮𝗹𝗹𝗼𝗰𝗮𝘁𝗶𝗼𝗻, 𝗱𝗮𝘁𝗮 - Fully enriched accounts - Signal-prioritized lists - Signal-validated targeting - Stronger ICP customer base - AI talk tracks per vertical - Enterprise fully resourced 𝗧𝗵𝗲 𝗿𝗲𝘀𝘂𝗹𝘁? 𝟮𝟱 𝗽𝗼𝗶𝗻𝘁 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗶𝗼𝗻 𝗹𝗶𝗳𝘁 The graphic below is the summary. Tomorrow's newsletter is the full breakdown. 𝗪𝗵𝗮𝘁'𝘀 𝗶𝗻 𝘁𝗵𝗲 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿 - How they validated which signals actually matter - 3 fixes that led to a 25-point conversion lift - Why they did all of this without a bigger budget If you're thinking about going upmarket, or you've started and the machine isn't keeping up, this one's for you. Subscribe to get it here: https://bit.ly/49RCm0h Or listen on the podcast "How ZoomInfo Re-Engineered GTM to Shift Upmarket" Spotify 👉 https://bit.ly/4b28Nwn Apple 👉 https://apple.co/3y8lch6 ✌️
7 Comments -
Danielle Bartone M.S.Ed
Palo Alto Networks is scaling fast, and they’re doing it with sales execution and data integrity moving in lockstep thanks to ZoomInfo. 🤝 From the sales floor to their data foundation, Palo Alto Networks relies on ZoomInfo to power smarter prospecting, clearer segmentation, and scalable GTM operations. The results? ✔️Increased sales productivity ✔️Greater operational efficiency ✔️Faster account cleanup at scale Watch this video and read the full story to learn more: https://bit.ly/4rXvVS5
-
Sahil Chawla
Hot take: if your “agent” is a 500‑line prompt wrapped in a webhook, you didn’t build an agent—you shipped a brittle script. The teams seeing step‑change gains aren’t asking one model to do everything. They’re assigning clear roles and letting multiple specialists work in parallel. Less monologue, more orchestra. Why swarms work: - Specialization: Smaller, focused contexts beat one mega‑prompt. - Parallelism: Research, drafting, and critique can run at once; results are reconciled, not hand‑waved. - Protocols over vibes: Tool use, interfaces, and checkpoints reduce fragility. Evidence and tooling: - Microsoft AutoGen shows multi‑agent collaboration improves coding/reasoning tasks: https://lnkd.in/gW_tyPBm - LangGraph enables stateful, graph‑based orchestration (with parallel branches): https://lnkd.in/gPPGXEt3 - OpenAI’s Swarm illustrates lightweight handoffs between agents: https://lnkd.in/gKeGPPVW - Research like CAMEL role‑playing (https://lnkd.in/g6VE6h4k) and Tree of Thoughts (https://lnkd.in/gYvFsS2W) underpin structured, multi‑step reasoning. For coding benchmarks see SWE‑bench: https://www.swebench.com/ A simple blueprint (R2P2): - Roles: Planner, Researcher, Builder, Reviewer, Operator. - Routing: A small orchestrator decides who acts next and with what inputs. - Parallelize: Fan‑out research/drafting; fan‑in aggregation and consensus. - Post‑checks: Tests, lint, fact‑checks, and human‑in‑the‑loop on low confidence. Practical patterns: - Sales intelligence: Parallel web/doc researchers produce briefs; a judge agent merges, de‑dupes, and cites. - Software changes: Test‑first planner → coder with tools → critic that runs tests → operator deploys behind a flag. - Support: Triage agent classifies; resolver drafts; QA agent verifies tone/facts; escalate on risk. Getting started this month: 1) Pick one painful, repeatable workflow. Write it as a checklist. 2) Turn steps into roles. Define inputs/outputs per role. 3) Orchestrate with a graph (LangGraph/Swarm). Add timeouts and retries. 4) Bake in evaluation (gold tasks, pass@k, latency, cost). 5) Log everything. Promote winning traces to playbooks. Prediction: The next wave is budget‑aware schedulers, event‑driven agents, and “agent ops” stacks for observability, policy, and safe tool execution. Single prompts won’t keep up. How are you structuring roles, routing, and review in your agent systems? What broke when you tried to scale a single agent—and what fixed it? Follow for more such content.#AI, #Business, #Innovation, #Tech, #Collaboration
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content