Vij Rajarajan

One line in this piece tends to surprise people, especially when it comes to system evaluation: “𝘐𝘧 𝘰𝘶𝘳 𝘣𝘦𝘯𝘤𝘩𝘮𝘢𝘳𝘬𝘴 𝘴𝘩𝘰𝘸 𝘮𝘰𝘳𝘦 𝘵𝘩𝘢𝘯 𝘢 50% 𝘴𝘶𝘤𝘤𝘦𝘴𝘴 𝘳𝘢𝘵𝘦, 𝘸𝘦’𝘳𝘦 𝘱𝘳𝘰𝘣𝘢𝘣𝘭𝘺 𝘯𝘰𝘵 𝘣𝘦𝘪𝘯𝘨 𝘢𝘮𝘣𝘪𝘵𝘪𝘰𝘶𝘴 𝘦𝘯𝘰𝘶𝘨𝘩.” That mindset grew out of many small failures and iterations. The point isn’t lowering the bar for quality. It’s raising the bar for how we measure systems and drive real improvements. This article reflects on that journey, from Flash Fill to today’s agentic systems, and why user flow, system evaluation, and human judgment matter in practice. Thanks to Anshika Mathews for capturing this so thoughtfully.

128

2 Comments

Vianai Systems, Inc.'s badass CTO, Navin Budhiraja, doing his thing. Insightful interview on the technological, professional and societal implications of AI and the need for enterprise grade AI technology to truly drive business performance.

2

Just wrapped the 7th Perplexity AI Business Fellowship session featuring Ali Ghodsi, CEO of Databricks, in conversation with Aravind Srinivas. Ali's refreshing candor about AI's enterprise impact, dropping some uncomfortable truths about enterprise AI that every engineering leader needs to hear: - "The only reason the AI bubble hasn't burst is inference-time compute" - not actual breakthroughs - When Databricks plugged advanced reasoning models into enterprise applications? "Barely any boost." The problem isn't intelligence but context - is "C" California or Celsius? - Success formula: unique data + distribution channels. Nothing else matters. - AI revolutionizes coding, which represents just 20% of engineering time. The other 80%? Human meetings and alignment work which AI can't touch. - Biggest blocker? Not capabilities but evaluation frameworks. "We have no way to measure reliability" in specialized enterprise contexts. This mirrors what we're seeing in healthcare AI - where reliable assessment frameworks are critical when patient outcomes are on the line. As I noted in my analysis of BVP's Healthcare AI Index (https://lnkd.in/gy3P7zdU), success isn't about algorithms but workflow integration and evaluation rigor. For engineering teams, the path forward isn't chasing fancier models but solving the semantic understanding and evaluation challenges that remain stubbornly human. #EnterpriseAI #EngineeringLeadership #HealthcareAI

24

Work 3.0: From Tool to Teammate — How AI is Rewriting the Org Chart and Redefining Productivity Last year, Aparna Chennapragada (Chief Product Officer at Microsoft) reminded us that organizations haven’t truly adopted AI , individuals have. Beneath the buzz, the real blockers are trust, organizational inertia, and how work itself is changing. (Link Below to 2024) This year at the MIT AI Conference, Aparna and I will unpack a couple themes: 🧠 How the AI revolution is moving from tools to teammates, fundamentally reshaping roles, teams, and workflows. 🤝 Why what’s magical today becomes tomorrow’s commodity . and how companies can build trust and stay ahead of the curve. ⚡ What it takes to build resilient, adaptive organizations in the age of intelligent systems. 📅 Join us for this high-impact conversation at the MIT AI Conference 🔗 https://lu.ma/mitai2025 We’re grateful to our sponsors: E14 Fund, Wilson Sonsini Goodrich & Rosati, OpenAI, MIT Sandbox Innovation Fund Program, HoundDog.ai, Dheyo AI, espressoai.it, BCG X, Egon Zehnder, Horizon Neuroscience, Roomie, TriNet, MIT Kuo Sharper Center for Prosperity and Entrepreneurship, Chonkie, Myndboosters AI, MemVerge, Cheyenne Ventures, Fenwick & West, Genspark. #MITAI2025 #AI #Innovation #Technology #FiresideChat #FutureOfWork #Work3 #AIProductivity #TrustInAI #Conference #Networking #MIT MIT Alumni Startups (MITAS): Shuja Keen, Daniela G., Adriana Catalina Vázquez Ortiz, Lorenzo Ligato, Dima Ayyash, Laura Hoang, Mainak Ghosh, Farzan Subhani, Nancy Wang, Serena Chung, Rand Hidayah, Rachel Lang, Luna Maroun, Zeenat Patrawala, Andreas Sanidiotis.

28

1 Comment

Excellent perspective from Jeetu Patel at Cisco. The point about trust is particularly important. One dimension that I believe deserves even more attention is the difference in speed between machines and humans. AI systems operate at machine speed, while governance, decision making, and oversight still largely happen at human speed. That gap creates a completely new category of risk. When AI systems are connected to critical infrastructure, financial systems, or core operational platforms, even a small error, misalignment, or manipulation can propagate at machine scale and in milliseconds. Humans often only realize what happened after the system has already executed thousands or millions of actions. That is why trust, guardrails, and strong oversight are so critical. The real challenge is not only building powerful AI systems but designing governance models that can keep pace with machine speed. As Professor Suraj Srinivasan from Harvard Business School recently put it: “the faster the car, the better the brakes needed”

1

Very happy to share that Vinay Rao is joining Samaipata as an Operating Partner. Vinay brings deep AI expertise built over decades at the frontier, including senior roles at Anthropic and Google, where he worked on real product and security challenges. He understands not just what these systems can do, but where they break. He is based in San Francisco and serves as CTO of ROOST.tools, a non-profit developing open source infrastructure aimed at improving safety standards across online platforms in the tech ecosystem. After joining our Annual Summit and leading a session on some of the toughest questions in AI today, including reliability, misuse and misalignment, it became clear that he could bring immense value to our community. We wanted to bring his expertise closer to our founders. As an Operating Partner, Vinay will mentor founders of our portfolio companies on critical decisions from day one. In early stage companies, choices around AI architecture, security and governance define speed, defensibility and long term risk. For our founders, having Vinay involved at those moments will make a real difference. Welcome to the team, Vinay. Excited for what is ahead! Thank you to Cinco Días for the coverage. More information in the link in the comments (in Spanish).

286

27 Comments

At AWS re:Invent — after an inspiring set of announcements from Matt Garman, Swami Sivasubramanian, and Ruba Borno discussions during Partner Advisory Board and executive sessions — I came away with a clear takeaway: AWS is accelerating its leadership in enterprise AI by delivering a comprehensive, end-to-end ecosystem for building, deploying, and scaling AI — not just in proof-of-concepts, but in production-grade, secure, and industry-specific applications. With new innovations such as NOVA 2, NOVA Forge, NOVA Act, KIRO Autonomus, Agent Frontier and AI Factory, AWS is providing everything customers — and their partners — need to successfully operationalize AI. In other words: AWS is handing you the full recipe to bake your desired “AI meal,” with proven ingredients, clear instructions, and safety guards baked in. This, I believe, represents a major leap forward for the “agentic AI” ecosystem. With these tools, organisations can finally transition from experimentation to scalable deployment — turning ideas from prototypes into real-world, impactful solutions. At Bexprt, we’ve set out from Day One with exactly this goal: to enable rapid, low-cost, secure development of industry-specific AI agents. The announcements at re:Invent bring us one step closer to realising that mission — and I’m excited to actively contribute and steer this transformation. These are truly revolutionary times. The future of enterprise AI looks promising indeed. Amazon Web Services (AWS) AWS Partners #AWS #AWSreinvent #AgenticAI #EnterpriseAI #AIFactory #NOVA #CloudInnovation #DigitalTransformation #AIatScale #TechLeadership

53

1 Comment

Did you hear it? … the song of the AI Agents dominating #awsreinvent2025 Nearly every other booth you walk to is demoing an #aiagent platform. In a nutshell, we’re talking about simple ai agents. I’d call them baby agents. A prompt, a RAG, and a few MCPs here and there to collect data, all governed by an agent studio to design agent workflows. Most platforms are promoting an open model framework where you can pick and choose your preferred LLM. Bedrock is one of the key options available. What’s missing? The business application. A lot of those platforms are in their very early stages, many are not even GA yet. What’s common: build from scratch mentality, and my ai agent platform can do everything, everywhere, all at once. IT buyers? Torn down between all of these platforms not knowing if it’s the right time to start choosing one or many. The potential? It’s going to be insanely huge, with businesses and organizations reaching incredible levels of automation, while also providing superior quality and experience to their customers. Get ready as 2026 will be exciting! [sounds surreal to say so, nearly have to apologize that my post did not use any single AI, not even for brainstorming or improving my style. WDYT: Better with or without AI?]

16

4 Comments