Activity
8K followers
-
Rajaram Srinivasan shared thisExcited to share that Unbound is a part of upcoming Google for Startups Gemini Startup Forum: Cybersecurity cohort - where 33 cybersecurity startups from around the world will team up with experts from Mandiant, WIZ, Chronicle, Google Cloud and Google DeepMind to fast-track growth, tackle real technical challenges, and explore what’s truly possible with Gemini. Can’t wait to connect with fellow founders shaping the future of AI this September at Google London. 🚀🚀🚀 Announcement post from Google in first comment below. Vince Bryant Google for Startups Google Cloud Security Google DeepMind
-
Rajaram Srinivasan shared thisAI has gotten so good that anyone can do anything now. A lot of people unfortunately think that the AI users with best output are the ones who use the latest skills, the newest plugins, or the trendy sub-agentic workflows. Most of that is noise. They are workflow optimizations at best. They doesn't objectively change the quality of the product you're building or the document you're writing. The real difference between someone whose AI output is excellent and someone with slop comes down to one thing: Do you know the job well enough to pressure test what AI gives you and catch where it's wrong? That step alone removes most of the slop, and what you ship starts looking like high-quality work. Last week someone shared a draft with me where the first paragraph had factual mistakes that made it obvious that they hadn't read the AI generated document once. Honestly, sending something like that is disrespectful. It means you asked AI for something, didn't bother to read it, and expected the person on the other end to do the due diligence for you. When anyone tries to sell you a course that's going to turn you into an AI engineer overnight, or AI-proof your career forever, just know that most of that is humbug. Today AI is at a point where thou shall ask and thou shall receive. The only work left is making sure what you received is what you asked for. It's that simple.
-
Rajaram Srinivasan shared thisJust back from the AI Agent Conference 2026 in New York, and the energy at our booth confirmed what we've been hearing on customer calls for months. Most conversations came back to the line we put on the standee: Don't secure agents. Secure actions. Static permissions and prompt-level guardrails break the second an agent does something autonomous. The builders and security leaders we met want runtime control over what AI coding agents can actually see, touch, and do and not just another policy deck. Picture shows Andrew Wesbecher in action! If you're shipping with AI coding agents and are worried about them accidentally wiping out your production database, ping us. If you're working on agent security and didn't catch us in NYC - let's talk. Unbound FirsthandVC
-
Rajaram Srinivasan shared this🚨 An AI agent just wiped a company’s production data - 𝐚𝐧𝐝 𝐢𝐭𝐬 𝐛𝐚𝐜𝐤𝐮𝐩𝐬. Not through some exotic exploit. Not through a malicious insider. But by using an API token it was never supposed to access. Here’s what happened: A Cursor agent, operating with scoped permissions, discovered a stray token buried in unrelated files. It used that token to execute a destructive command, permanently deleting critical infrastructure. 𝐍𝐨𝐭𝐡𝐢𝐧𝐠 𝐬𝐭𝐨𝐩𝐩𝐞𝐝 𝐢𝐭. 𝐇𝐞𝐫𝐞’𝐬 𝐭𝐡𝐞 𝐮𝐧𝐜𝐨𝐦𝐟𝐨𝐫𝐭𝐚𝐛𝐥𝐞 𝐭𝐫𝐮𝐭𝐡: 𝐀𝐈 𝐬𝐞𝐜𝐮𝐫𝐢𝐭𝐲 𝐩𝐨𝐬𝐭𝐮𝐫𝐞 ≠ 𝐬𝐞𝐜𝐮𝐫𝐢𝐭𝐲 You can: • Discover every agent in your environment • Lock down permissions for each sub-agent • Ship “security skill files” that say: pretty please, with a cherry-on-top, don't delete my prod database …and still get taken down. 𝐖𝐡𝐲? Because none of that is enforcement. Let’s be precise: 𝐑𝐮𝐥𝐞𝐬 ≠ 𝐆𝐮𝐚𝐫𝐝𝐫𝐚𝐢𝐥𝐬 The agent knew the rules — and acted anyway. 𝐏𝐞𝐫𝐦𝐢𝐬𝐬𝐢𝐨𝐧𝐬 ≠ 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐦𝐞𝐧𝐭 It found another path. Just like a capable human would. 𝐒𝐞𝐜𝐮𝐫𝐢𝐭𝐲 𝐩𝐨𝐬𝐭𝐮𝐫𝐞 ≠ 𝐂𝐨𝐧𝐭𝐫𝐨𝐥 Visibility is not prevention. The only thing that works is 𝐝𝐞𝐭𝐞𝐫𝐦𝐢𝐧𝐢𝐬𝐭𝐢𝐜 𝐠𝐮𝐚𝐫𝐝𝐫𝐚𝐢𝐥𝐬: 👉 Intercept actions before execution 👉 Treat every tool call as a hard control point — not a suggestion That means using 𝐧𝐚𝐭𝐢𝐯𝐞 𝐡𝐨𝐨𝐤𝐬 𝐛𝐞𝐟𝐨𝐫𝐞 𝐭𝐨𝐨𝐥 𝐮𝐬𝐞 to block unsafe actions outright. We used to call this “Phase 3 maturity” in Unbound docs. Well, we we wrong. 𝐓𝐡𝐢𝐬 𝐢𝐬 𝐭𝐚𝐛𝐥𝐞 𝐬𝐭𝐚𝐤𝐞𝐬 𝐧𝐨𝐰. At Unbound, we’re focused on exactly this problem: bringing real control into the execution layer, where it actually matters.
-
Rajaram Srinivasan shared thisWe are starting to see a weird new pattern of usage on AI coding agents. Companies buy Claude Code for enterprise. Teams burn through usage limits early in the month. Developers still need to ship. So they switch to personal Claude Max and keep going. No drama. No malicious intent. Just a very practical workaround. So common that it's easy to overlook. What's worse? the local setup still feels exactly the same. Many skills remain available. Some MCP servers may still be configured, easy to re-authenticate if not. The codebase is still on the machine. The agent still has access to local context. But the account boundary has changed. Now your enterprise workflow may be running through a consumer account: Outside your admin controls. Outside your retention policies. Outside your audit visibility. Outside your enterprise data protections. Purchasing an enterprise plan is not the end of securing claude code, it's the beginning of it. Security controls have to follow the agent. What can it see, touch and do? Of course, this is one of the things we solve for at Unbound. if you want to discover and secure coding assistants, their supply chain (MCP, skills, plugins) and their actions (terminal command runs, MCP tool use etc.), drop me a note. Happy to chat. #ClaudeCode #Anthropic #AISecurity #AgenticAI #MCP #Cybersecurity
-
Rajaram Srinivasan shared thisFrom Microsoft free tier to Microsoft for Startups spotlight. Pretty surreal to see Unbound featured by Microsoft for Startups alongside companies like Pinecone and Weights & Biases. Microsoft was the first cloud platform we signed up for in the early days of Unbound. We had no external funding, were trying to make every dollar stretch, and that generous free tier gave us real room to build. Grateful for the support then, and honored to be featured now. Still early. Still building. Link in first comment for upto $150k in credits. #Startups #MicrosoftForStartups #CyberSecurity #AI #Founders
-
Rajaram Srinivasan shared thisThe scariest part of AI coding agents is not that they can write code. It is that they can act. Most security teams are still treating this like a policy problem. It isn’t. AI coding agents do not just generate code. They make decisions. They execute commands. They modify files. They restart services. And humans (if they're reviewing) are often approving any action it suggests casually. The risk is no longer just bad generation. It is ungoverned agency. This is exactly why “allow or block” policy controls is starting to break down. Unbound partnered with Google Cloud Security on a new whitepaper, Securing AI Coding Agents, that lays out what a continuous governance framework actually looks like in practice - one that gets smarter every week, not every quarter. I'll be at Google Cloud Next '26 next week. If you're thinking about how to scale coding agents without losing control, come say hello.
-
Rajaram Srinivasan shared this#RSAC 2026 is one for the books. We kicked things off with a pre-RSA dinner co-hosted with Kabir Mathur and Zain Rizavi, then stayed fully booked every day from breakfast to dinner with nonstop conversations at the booth and beyond. Walking out of RSA with weeks of packed calendars - exactly as it should be. Shoutout to the team that made this week what it was - Vignesh Subbiah, Andrew Wesbecher, and Tyler Hayden! Now, finally, enjoying the calm after the storm.
-
Rajaram Srinivasan shared thisThis #RSAC, every vendor on the expo floor is going to be selling AI security or AI for security. So here’s a simple field guide for navigating the noise: Ask every vendor one hard question - What exact AI security problem do you solve? Make them answer precisely. Is it security for: customer-facing AI applications? employee use of AI tools? coding agents? browser agents? MCP servers and tool access? model behavior? red teaming? prompt injection? data leakage? agent action controls? Because these are not the same problem. And here’s the easiest tell: If a startup says “we do AI red teaming” and “we do MCP security” in the same pitch, you already have your answer. That is not a unified product. That is confused positioning wrapped in a hot market. This market is already splitting into real camps. One camp is securing AI products companies ship to customers. Another is securing employee use of AI inside the enterprise. Another is emerging around agent access, tool use, and action control. Trying to be all of them at once is like a startup claiming it will be both the Wiz and the Zscaler of AI. Yes, building is easier now, and funding rounds are bigger, but great companies still win by solving one narrow, painful problem exceptionally well before becoming a platform. In early stages, breadth is too often just a cover for lack of depth. So, ask your vendor which is that lane that they're going deep on. I’ll go first: We’re building Unbound - the industry-first AI Agent Access Security Broker. We secure employee usage of AI coding agents by ensuring agentic actions align with requested user behavior. For example, we’d stop Kiro from deleting and recreating a production environment and causing a 13-hour outage. Blog in comments.
-
Rajaram Srinivasan liked thisRajaram Srinivasan liked thisWe just wrapped up a huge hiring push at Clearly AI. After reviewing hundreds of applications, we're excited to welcome six incredible new teammates (more on them soon!) Reviewing candidates across sales, product, engineering, and support, I kept seeing the same patterns, so I turned the lessons into a practical guide for anyone who's interviewing: https://lnkd.in/g-4xqSrM I hope it's useful.Getting Hired: What I wish every candidate knew - Emily Choi-GreeneGetting Hired: What I wish every candidate knew - Emily Choi-Greene
-
Rajaram Srinivasan liked thisRajaram Srinivasan liked thisI am proud to announce that Xactly Corp has been recognized as a Leader in the 2026 Gartner® Magic Quadrant™ for Sales Performance Management. This achievement reflects our team’s commitment to developing an intelligent platform that transforms how companies drive revenue. Congratulations to the entire Xactly team for their dedication and hard work! This recognition signifies not only our current position but also the ongoing efforts we are making to innovate and create real impact for our customers. We are just getting started. Read more about this recognition here: https://lnkd.in/gBZe7Zvt #Xactly #GartnerMagicQuadrant #SalesPerformanceManagement #SalesPerformanceOrchestration2026 Gartner® Magic Quadrant: Xactly Named a Leader in Sales Performance Management2026 Gartner® Magic Quadrant: Xactly Named a Leader in Sales Performance Management
-
Rajaram Srinivasan liked thisRajaram Srinivasan liked thisGroww’s testimonial is a special one for us. They were one of the early believers in what we were building at Matters... As their cloud and data footprint expanded, the challenge was not just discovering sensitive data. It was reducing alert fatigue, simplifying security operations, and moving from reactive investigations to continuous remediation, without slowing down engineering or production. By bringing fragmented security signals together with deeper context and intelligent remediation, Matters helped the team identify risks earlier and act on them continuously. This is the value we want to bring to every enterprise: stronger security, lower operational overhead, and the confidence to scale without friction. Grateful to Prajal Kulkarni, Pradeep Bhat, and the entire Groww security team for their early trust and continued partnership. Video below. Full case study in the comments.
-
Rajaram Srinivasan liked thisRajaram Srinivasan liked thisIntroducing AI-10 Labs 🚀 . "Early Stage AI is Booming. Not Slowing"; "-1 to 1 and beyond" — the journey continues. Special Panel: "Building AI-Native Startups on Shifting Sands 🚀" Early preview: June 17th | 6–8pm | Palo Alto, CA Hosted at Fiddler AI offices. 550 California Ave, Suite 200. Limited to 30. Super thrilled to introduce "AI-10 Labs" & our thesis on "early stage AI-native startup" opportunity — to select entrepreneurs, investors, and industry executives. Tremendous excitement and heartfelt gratitude to Dayakar Puskoor and Dallas Venture Capital team for all the support and partnership on this early exploration. Agentic AI is a generational opportunity — yet it must be harnessed carefully in the NEW "shifting sands" paradigm. Super thrilled that Doug Dooley (COO, Data Theorem, x-venrock), Krishna Gade (Founder & CEO, Fiddler AI), and Sanjay Kalra (Founder & CEO, Stealth AI-Cyber) will share their practical insights on the critical topics on everyone'e mind. * Frontier models unleashing huge leaps — amid scrutiny, policy shifts, and new governance and sovereign controls. * Rapidly emerging new agent harnesses * Edge & endpoint intelligence emerging for better control and cost. * Can we imagine a "gateway free" - no chokepoints - for Intelligence Cyber. * Agent drift requiring need for runtime AI governance beyond observability. * Token costs forcing serious guardrails on spend and ROI. And the holy grail: Is this a "Too Big to Fail" opportunity — only for those with deep pockets and gigantic valuations? How do we democratize the pace and opportunity — and make it sustainable? *** Doug, Krishna, Sanjay — grateful for your time and practical wisdom; and our trusted partnership over the years. Krishna and the Fiddler team — thank you for graciously hosting us. 🙏 To all the AI pioneers - founders, investors, and passionate players in the industry— thank you for the inspiration to try this "-1 to 1" act. 🚀 🚀 🚀 RSVP link in comments. Seats are limited. #AI10Labs #EarlyStageAI #AgenticAI #ShiftingSands #DVC #dallasventurecapital #-1to1
-
Rajaram Srinivasan liked thisNot to bury the lede on this, but the key to this kind of ongoing acceleration at Benchling isn't a choice of speed over quality, it's our intentional focus on leveraging AI to improve both speed AND quality. "Shipping faster doesn't mean shipping with less scrutiny. Every pull request from this sprint went through Benchling's standard review process, including the human sign-off, testing requirements, and automated assurance checks that govern every release. The AI agents handled scoping and implementation. Engineers stayed accountable for what merged."Rajaram Srinivasan liked this💬 You asked. So we built it. Sometimes it’s the smallest improvements that can have the biggest impact on your day. That’s why every year, our product and engineering teams dedicate a week to tackling the small but meaningful fixes our customers have been asking for. This year, we received 541 ideas from scientists, admins, and lab leaders — and shipped 114 improvements in just five days.
-
Rajaram Srinivasan liked thisRajaram Srinivasan liked thisHappy Thursday, here are the hottest startups hiring this week: 🌱 𝗣𝗿𝗲-𝘀𝗲𝗲𝗱 / 𝗦𝗲𝗲𝗱 Structured AI (YC F25) (Led by Raymond Zhao, Issy Greenslade, Brandon Abreu Smith) - Hiring for Engineering, Operations, GTM Lua AI (YC F25) (Led by Stefan Kruger, Lorcan O Cathain) - Hiring for Solutions Architect, Implementation MSPilot (YC F25) (Led by James Jiang) - Hiring for Engineering, GTM Unbound (YC S24) (Led by Rajaram Srinivasan, Vignesh Subbiah) - Hiring for Growth, Engineering CollectWise (YC F24) (Led by Sean O'Brien) - Hiring for GTM AgentPhone (YC P26) (Led by Manav Modi, Meet Modi) - Hiring for Engineering 🚀 𝗦𝗲𝗿𝗶𝗲𝘀 𝗔 / 𝗕 Spade (YC W22) (Led by Oban MacTavish) - Hiring for Engineering, Finance BlueCargo (YC S18) (Led by Alexandra Griffon, Laura Theveniau) - Hiring for GTM HavocAI (Led by Paul Lwin, Joe Turner) - Hiring for Engineering, IT, Program Composio (Led by Soham Ganatra (Hiring), Karan Vaidya) - Hiring for Engineering Cafler (Led by Ricard Guillem, Iñigo Diego Cavero) - Hiring for Operations, Product, GTM 📈 𝗦𝗲𝗿𝗶𝗲𝘀 𝗖+ Quince (Led by Sourabh Mahajan, Sid G.) - Hiring for Product, Operations, Supply Chain, Marketing, GTM ScaleOps (Led by Yodar Shafrir) - Hiring for Engineering, Operations, Marketing, GTM Follow me for more startup job opportunities updated weekly! #startups #hiring #careers #jobsearch #LOCKIN
-
Rajaram Srinivasan liked thisCISO360 Global Congress “Different industries, different regulations, different perspectives—but many of the same questions, opportunities, and many of the same risks.” I decided to try something new: an open dialogue with the audience co-hosted with Ray Stanton. My goal was simple: make the audience part of the conversation and ensure everyone walked away with practical insights they could apply immediately. Thank you to everyone who participated, challenged our thinking, and shared your perspectives. These conversations remind me that while AI is evolving at an incredible pace, our greatest advantage is learning from one another. Looking forward to continuing these discussions around the world. #ArtificialIntelligence #AISecurity #CyberSecurity #Leadership #Innovation #Community #AIGovernance
Experience
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Education
-
Massachusetts Institute of Technology
-
-
Activities and Societies: Martin Trust Center for MIT Entrepreneurship, MIT Sandbox, Marketing, Company relations at Product Management Club, Member at Design Club, Leadership Club and Systems Thinking Club, Entrepreneurship and Innovation Club
-
-
-
Volunteer Experience
-
Mentor
Technovation Challenge
- 4 months
Education
-
Recommendations received
2 people have recommended Rajaram
Join now to viewOther similar profiles
Explore more posts
-
Amanda Herson
The best product recs come from the front lines. This week, we hosted a forum of top consumer founders in NYC, from a brand launched last month to a company doing 9 figures in revenue. Here’s what they cited as their favorite AI tools right now: 👑 Claude is consensus. 1. Claude: https://claude.ai ➤ The #1 answer by a mile, used for everything from gathering simulated feedback from customer personas to building custom iMessage apps. Claude for Excel earned its own shout-out for financial modeling. 🎙️ Voice-to-text is popular but fragmented. 2. Wispr Flow (mentioned 2x): https://wisprflow.ai/ 3. Willow: https://willowvoice.com/ 4. Monologue: https://www.monologue.to/ ⚡ Productivity picks are scattered but with glowing reviews. 5. Hyperagent by Airtable (in alpha): https://lnkd.in/eUPnMNtU ➤ Platform for using and deploying ‘AGI-level’ agents in the cloud (our team at FC is testing and hugely impressed) 6. Lightfern: https://lightfern.com/ ➤ Gmail autocomplete tool (reportedly blows every competitor out of the water 👀) 7. Cursor: https://cursor.com/ ➤ Coding copilot, helpful for rapid prototyping 8. Gemini by Google DeepMind: https://gemini.google.com/ ➤ Particularly because of its Google data integration 🎨 Favored creative tools are multimedia. 9. Suno: https://suno.com/ ➤ Music generation 10. Midjourney: https://lnkd.in/eAWNzTsX ➤ Image generation, but shouted out for 3D animations 11. Untold: https://www.untoldapp.com/ ➤ Journaling 🧠 Bonus recs from the group - 2 thought leaders worth following: 12. How I AI by Claire Vo on YouTube (or wherever you listen to podcasts - but the screensharing is gold): https://lnkd.in/eNj6jb6n 13. Aakash Gupta on X: https://x.com/aakashgupta Hope a few of these are new to you. Was your favorite mentioned?
9 Comments -
Hadley Harris
In my experience, ~17 people is the tipping point where an org stops operating as a single atomic unit and starts fragmenting. Fragmentation kills speed. AI lets you push scale without pushing headcount. With all the tooling we're building, I don't see why Eniac Ventures would ever need to exceed 17 That's the future of VC
10 Comments -
David George
I had a great time chatting with Patrick O'Shaughnessy on Invest Like The Best. I've known Patrick since college, and this is the first time we've talked markets and investing at this much depth. The fundamentals of company building haven’t changed: people, products, and markets matter. But obviously, private markets have evolved substantially over my career: there are now ~6x more private unicorns than public companies with a $1b+ market cap. And at the end of 2010, just 2 public technology companies were among the top 10 in market cap; today it’s 8 of 10. AI (alongside software eating everything more generally) is clearly driving a lot of this. But it’s instructive to look at everything from the steam engine, to the early days of Facebook and Google user monetization, to real-time success stories like Databricks, Anduril, OpenAI and Waymo, to get a clear picture of where the opportunities lie. It was a pleasure to go deep on all this and more!
9 Comments -
Kit Yu
There is a scenario where SaaS vendors benefit from Cowork adoption as a generalist intelligence layer that pulls through domain-specific AI. As an example, on 2/24, Intuit announced a partnership with Anthropic. Intuit will gain access to Anthropic’s models and custom AI agents within its businesses via the Claude Agent SDK, and Anthropic will feature Intuit’s products on the Cowork, Claude for Enterprise, and Claude.ai platforms through MCP integrations. We view this as a move to amplify the strengths of both platforms noting that the integration/agent orchestration framework in this partnership cuts both ways. On the one hand, Intuit gave the example of a restaurant group with 15 locations who will be able to prompt Intuit Intelligence to orchestrate and deploy Claude within IES. On the other, a solopreneur can connect a spreadsheet of their transactions to Claude and use Intuit’s invoicing/payments tools within the Claude app to generate invoices. The financial terms of the contract have not been disclosed but for context, this announcement followed Intuit’s $100mn/year deal with OpenAI and we believe the logic could be similar: the foundation models do not need to build custom workflows and instead can monetize their models and reach, while Intuit maintains control of the data and preserves their unit economics. Ultimately, adding a layer between the customer and Intuit could create risks of value abstraction in the long run. That said, the agreement could drive better customer acquisition, akin to a partnership with a search engine in the Web 2.0 era. The bull case is that Intuit can roll out and drive adoption of their own agents, leveraging Anthropic’s models and interfaces to deliver better outcomes for customers than either party could have alone.
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top contentOthers named Rajaram Srinivasan in United States
5 others named Rajaram Srinivasan in United States are on LinkedIn
See others named Rajaram Srinivasan