About
Building Calibre, an applied AI Research and Consulting firm that helps CXOs and product…
Articles by Sandhya
Activity
19K followers
-
Sandhya Hegde shared thisI work with talented agent engineering teams at AI startups as well as leading public software cos and the #1 most under-rated investment is having a well-defined #Agent_Success_Rate. Without this composite north star metric for every agent session, it's impossible to hill climb agent quality systematically. Teams can't get immediate feedback on new releases or communicate their progress to leaders easily. They rely on lagging business metrics that are influenced by factors outside of agent quality. This is a huge reason why customer support and coding agents have flourished - they have been able to hill climb agent success metrics like ticket resolution and code acceptance. But many other domains don't have these obvious metrics - your product team needs to come up with one! On June 9th, Justin Bauer and I sharing the Calibre framework for defining Agent Success Rate on June 9th as part of Amplitude's free AI Analytics for Builders series on Maven. Sign up to attend live or get the recording: https://lnkd.in/dgauFikS
-
Sandhya Hegde shared this😓The Friction is your Judgement Loved this AIEng Europe talk. Friction is not only how you express judgement but also how you learn and build taste! Justin Bauer and work with many product teams on improving quality with evals and monitoring. We always see builders on the bleeding edge eager to automate away the whole process before they understand it. But you can’t express taste and judgement without first putting in the manual work and creating the ground truth that cloud agents can use to automatically improve your systems. No pain, no gain. https://lnkd.in/g79B4fETThe Friction is Your Judgment — Armin Ronacher & Cristina Poncela Cubeiro, EarendilThe Friction is Your Judgment — Armin Ronacher & Cristina Poncela Cubeiro, Earendil
-
Sandhya Hegde posted thisThe case for demoting yourself. CEOs are incredibly frustrated with their managers right now. As a manager/leader, one of the most important roles you have is to define a clear picture of what success looks like for your team. This is hard to do when what's considered good in those roles has fundamentally changed and you have never seen or done this new job yourself. Ever. There are two options: going back to the builder path or blowing up your manager path. If you are managing a relatively small team and always considered yourself more of a builder than a manager - embrace the super IC path and reinvent yourself. Even when I had a team of 60+ people at a startup, I considered myself mostly a great IC+mediocre manager and tried to lead by example. If you are like that, embrace your craft. The career opportunities for AI builders with judgement and security awareness are boundless (and lucrative). If you are not really a builder, consider the fact that the skills that served you well in the past (managing up, getting headcount, navigating politics and egos) are not going to be as valuable in tech going forward. Stop asking for headcount and disrupt your own team. Have 15 direct reports instead of 5. If you were the VP before, you are now also the Directors. Let go of ICs who are not reinventing themselves. Hire fresh AI-native talent. It might sound counter intuitive but demoting yourself might be the best thing you can do right now. Consider doing it before it is done to you by the CFO's new budget.
-
Sandhya Hegde posted thisWatching the Cowork wave flood into GTM and Ops teams in big companies and dear lord, it has never been more valuable to understand how software works. Creative system thinkers are on a high.... but folks without an intuition for software are building skills and plugins the wrong way - not able to make any edge cases work, building fragile, slow, token-expensive solutions, not able to test their own skills.. My big takeaway from this: 🤖 Either AGI will correct people towards better approaches to achieve their goals or 💰There will always be a massive market for vertical agent platforms that abstract the underlying systems and best practices so people don't need to know how software works
-
Sandhya Hegde shared this💀 How to survive the death of SaaS: if you are still debating “should we build for agents?”, you need to immediately shift the conversation to “what will it take to ship great Agent Experience?” TLDR; 🤖 full capability parity between what humans and agents can do 🚀 CLI/SDK/APIs for agents to provision and configure your product 🎓 evolving skills that encode your best practitioner judgment 💰high performance low cost vertical models built from open source LLMs 🪨 stable interfaces and harnesses that outlast model changes Full article in comments:
-
Sandhya Hegde shared thisWe're kicking off our first live cohort on Reforge focused on AI Evaluation for agentic product development tomorrow with >1000 PMs and I've been reflecting on how much has changed in just the last 3 months.. ◼️ With coding agents getting good, literally everyone on the team should be contributing to automating evals - not just the AI engineers ◼️ You can write evals to inspect and improve anything - from skill files that automate your PRD and review process to the product development lifecycle itself ◼️ Teams that took the time to build a strong foundation last year feel like they are flying this year - products getting better every week, with automated monitors scoring user experiences and fixing bugs What has NOT changed is this - too few people do the hard manual work of building the foundation, actually looking at trace data, building stronger datasets and simulations, and aligning their automated evals to their team's taste. If we can change that, Calibre succeeds in its mission. Wish us luck for a great day 1 workshop tomorrow ☺️
-
Sandhya Hegde shared thisIn a feat literally no one predicted a mere 6 months ago, Anthropic just overtook OpenAI in annual run rate revenue after spending 3X less on training models. A case study for the ages on business focus, how constraints lead to better innovation and the second mover advantage 🙌🙌🙌 📈 Anthropic: 10× growth/year — $1B → $30B in ~16 months 📊 OpenAI: ~$27B estimated end of March at a 3.4×/year trajectory 💼 Enterprise LLM API share: Anthropic 32% vs OpenAI 25% 🤖 Claude Code alone: over $5B ARR in under 10 months - could become the first product in the history of the universe to hit $10B ARR in its first year come May 🤯 PS: when you have such extraordinary product market fit, it's not the 10 cool tricks by the growth team that shapes trajectory
-
Sandhya Hegde posted thisRemember when they said every expert would have an AI that people would want to interact with? Turns out that's just a good Skills.md file. For our agents to read.
-
Sandhya Hegde shared thisThe most underrated AI risk I see in software companies right now isn't growth rate or efficiency. It's talent gravity: your sheer ability to keep great people and attract more of them. The good people are now amazing. But the bad are terrible. They are either stagnant or shipping slop and calling meetings with everyone else to review poor work. Their coworkers dread having to collaborate with them. Same is true for companies - AI has made it obvious who is reinventing themselves and accelerating vs stagnating. Blunt tools like AI mandates and widespread layoffs won't fix talent gravity. What will is harder. Wrote up how to think about team reinvention inspired by my recent conversations with Spenser Skates and a great CXO dinner hosted by Alex Bilmes 🙌
Sandhya Hegde commented on a post
2mo
❤️ 🌊 can't wait to try this... It is so perfectly Amplitude's mission and the timing couldn't be better - yes loops, no slop.
Product data is the only way to push back on the deluge of slop code and slop features inundating all of us right now. It's great that agentic coding is "solved" but we really really need agentic "product" to work.
-
Sandhya Hegde reacted on thisSandhya Hegde reacted on thisNot dead yet! Always wonderful to see reaccelerating growth.
-
Sandhya Hegde reacted on thisSandhya Hegde reacted on thisThe last few years have been hard, especially Q3 2024 when our growth rate had bottomed to 6.5%. It's been a slow but steady climb since, and we finally crossed 20% again. We are going to keep accelerating our growth the same way we have over the last couple of years. We had a strong Q2, results below: - Second quarter revenue of $100.9 million, up 21% year-over-year - Annual Recurring Revenue of $410 million, up 22% year-over-year - Remaining Performance Obligations of $483 million, up 35% year-over-year - Second quarter net cash provided by operations of $25.6 million and Free Cash Flow of $23.7 million Amplitude Q2 earnings: https://lnkd.in/gBccCk_p Chart source: https://lnkd.in/gXvV6Rmx
-
Sandhya Hegde reacted on thisSandhya Hegde reacted on thisToday, iMerit officially became an EXL company. EXL is a $2.5 billion, NASDAQ-listed data and AI company, led by Rohit Kapoor and his incredible leadership team that has stayed together and grown the company for two decades. EXL brings deep enterprise knowledge, global scale, and AI platforms built for business-critical operations. On behalf of iMerit, I am proud and delighted to become part of this exceptional combination uniting EXL's extraordinary depth in domain, data, and enterprise transformation with iMerit's work in AI models, evaluation, and human intelligence. Thank you Rohit for your warm welcome. I am energized by what lies ahead and honored to join EXL’s executive committee. Together, we will help enterprises build, fine-tune, and operationalize domain-specific AI while continuing to serve the exacting needs of frontier labs. When I signed the agreement in the EXL New York office, my mind went back to one small team in the Sunderbans, where iMerit started. Twelve years and thousands of careers connect these two places. Along the way, our teams earned the trust of pioneers in autonomous mobility, healthcare, hi-tech and robotics. We built deep expertise in computer vision, generative AI, and multimodal AI applications. Today, we have a ringside seat in the development of AI, working with frontier labs, AI-native innovators and enterprises to build AI systems for real-world use. The stakes have gone from “can we annotate an image?” to “can we help a car make the right decision in the rain; can a model flag a tumor reliably in a scan?” That's why we've always believed better-quality data, not more data, is the future of AI and why we call expert data the third pillar of AI. What matters most to me is how we arrived here. We kept two feet firmly on the ground: one for our customers and one for our employees. Everything iMerit is, stands on that balance. That belief is also reflected in the teams we have built. People have often asked how iMerit achieved an even gender split as an AI company. The answer is simple: the world is 50-50, so our company naturally reflects it. To our clients and partners: the builders, the frontier labs, the enterprises who trusted us with their hardest problems: our deepest gratitude. To our investors and board, we are privileged to have such ardent sponsors and believers in our work. Thousands of our employees and their families, the hands and hearts that have built iMerit. are economically empowered through this journey and we have also supported two foundations that we deeply believe in and partner with. This moment belongs to all of us! Today iMerit, an EXL company, is ready for what's next. To become part of a much bigger journey - with a deep commitment to clients, to EXL and with the same two feet on the ground. And still, at this juncture, I remain unmoved in the belief that true breakthroughs happen at the intersection of people and machines. Press release is now live. Link in comments.
-
Sandhya Hegde reacted on thisSandhya Hegde reacted on thisToday, I'm excited to announce the launch of Multi-Model. Token spend is the new headcount problem. Are we getting enough out of what we're spending? The team just shipped Multi-Model to help answer that. Many of the marketing leaders I've been speaking to this year are running every Playbook through the same expensive model. Competitive research, meta descriptions, campaign reporting. All the same tier, regardless of what the job actually needs. That's the wrong tradeoff. Deep research on a competitive teardown needs frontier intelligence, but a meta description doesn't. Get the match wrong, and you're either overpaying for simple work or asking a cheap model to carry something it can't. Multi-Model gives you control and reasoning efforts over which specific models to leverage, for different Playbook outputs. Alternatively, hand it to Auto Mode and let us pick. Quality doesn't come from the model alone. It comes from the harness around it, the research, brand grounding, and citation checks built specifically for marketing work. Not to mention the marketer in the loop. That's what holds voice and accuracy together when you change the model underneath. Read more about it here: https://lnkd.in/gFUPdWGS
-
Sandhya Hegde reacted on thisSandhya Hegde reacted on thisHow to do customer marketing in 7 seconds: 1. Stumble on a nice customer quote 2. Screenshot it 3. Drop it into a ploy 4. publish My fav page: https://lnkd.in/ggcfT6sf. thanks Matthew Fine for my LI content of the day!
-
Sandhya Hegde reacted on thisSandhya Hegde reacted on thisThe last GUI you'll ever build? That was the provocative question Justin Bauer and Sandhya Hegde from Calibre Labs challenged us with last week during our AI Accelerators session at Docusign. A few ideas have been sticking with me. The way we build software is shifting from interfaces to intelligence. Instead of starting with the UI, AI-native products are being built around tools, reasoning, and orchestration, with the interface becoming just one layer of the experience. A few takeaways that resonated: • Build from the inside out. Start with MCP tools and capabilities before designing the GUI. • Agent builders aren't workflow builders. Treating them like drag-and-drop automation misses the point—reasoning and adaptability are the product. • Evaluation is a first-class product discipline. Shipping without robust eval datasets is like launching without tests. Perhaps the biggest takeaway wasn't about technology—it was about how product teams are evolving. • PMs are writing evals and shaping agent behavior. Designers are optimizing for outcomes instead of screens. Engineers are thinking about global agent layers, governance, and orchestration as core architecture decisions. • As enterprise AI matures, governance, trust, and evaluation are becoming just as important as model quality. Someone asked me after the talk - So how does it feel to build the last GUI? :) Thanks Justin and Sandhya for an insightful discussion and for challenging us to rethink how we build AI-native products. Excited to see where this next generation of software takes us. #AI #ProductManagement #AINative #ProductDevelopment #EnterpriseAI #Agents
-
Sandhya Hegde reacted on thisSandhya Hegde reacted on thisToday, I am excited to announce Stigg 2.0. But first let me tell you how we got here 👇 Every AI company is accidentally building a bank. Credit wallets with prepaid balances. Double-entry ledgers. Reserve-and-settle mechanics for long-running agents. Priority-based consumption rules. Balance checks on every API call before work begins. These are financial primitives - and almost every AI company is rebuilding them from scratch. Four years ago, Anton and I started Stigg with a bet: real-time entitlements would become one of the most critical infrastructure layers in modern software. We built that engine. Webflow, Qlik, Miro, Upwork, PagerDuty, and many others rely on it in production. We were right about the architecture. The market just wasn't ready yet. Then four forces converged at once: AI companies started shipping faster than their pricing infrastructure could keep up. Agentic workloads demanded millisecond-latency enforcement - not eventual consistency. Enterprise buyers started blocking AI deals without usage governance. And OpenAI published Beyond Rate Limits, describing the same synchronous decision engine we'd already built - concluding no third-party vendor could provide it. We read that post and recognized our own architecture. So, we are announcing today the usage runtime for AI products. Not billing. Not metering. The real-time enforcement and governance layer that decides what every customer, user, team, and agent is allowed to do - at the moment they try to do it. Billing tells you what happened. Stigg decides what's allowed. Here's what we shipped: 1/ A financial-grade credits engine. Double-entry ledger, zero-overdraft enforcement, priority-based consumption, ASC 606 audit trail. Credits done right. 2/ The first millisecond-latency AI usage governance layer. Per-user, per-team, per-agent budget controls enforced at consumption - not on the invoice. Enterprise buyers need this to sign. Just like SSO before, enterprise-ready in AI means usage governance. 3/ BYOC Usage Metering - deployed into your VPC. Scales to +1M events/sec. Usage events never leave your network. Fixed fee - no more per event pricing! 4/ We acquired Received.ai - its founding team has joined Stigg. Now fully integrated: bespoke multi-year SLG contracts with custom pricing, tiered commitments, prepaid credit pools, and billing schedules that don't fit in a dashboard. 5/ Agentic tooling, MCP server, CLI-first operations. 6/ New Pricing The full deep-dive - the architecture, the market forces, and why we believe entitlements are the abstraction that changes everything - is in the blog linked in comments. We're at AI Engineer World's Fair. Booth L G03. Come break our demos. To the Stigg team - you rebuilt everything. New engine. New governance product. New BYOC architecture. New website. New pricing. What you pulled off is extraordinary. To the customers who trusted us before the world caught up: you shaped this more than you know.
-
Sandhya Hegde reacted on thisSandhya Hegde reacted on thisToday marks a major milestone in iMerit Technology’s relentless journey. Data and AI leader EXL announced that it has entered into a definitive agreement to acquire iMerit. This is indeed a historic day for us. iMerit ethos is two feet planted firmly on the ground – one for customers & one for employees. We have flourished by paying attention to both of those in the AI industry. We have enjoyed working with all the pioneers in AI across autonomous mobility, healthcare, GenAI & now robotics. We grew our leadership in Computer Vision & now we are now working with the top AI labs who are defining the future of multimodal AI. Meanwhile, the industry has reached a point where the AI experimentation needs to move to AI production at a massive scale. AI must be deployed, tuned, & managed inside thousands of large companies across sectors like insurance, automotive, finance, banking & healthcare. These companies need a trusted partner who understands their business & understands how to bring cutting edge AI in a safe, accurate, & trusted manner, into their workflows. This is why we are joining hands with EXL. Both companies share a belief that specialized high-quality data is the foundation of AI success. Both companies believe in the sanctity of customer data. EXL has built a great reputation in those industries & has a deep understanding of their business workflows. Their customers trust them to implement AI in & tune it on their proprietary data by leveraging iMerit’s expertise. Together, we can be a force in the market with increased capabilities & scale the business. In this moment, I want to acknowledge how far we have come in these 10+ years of being in AI. Along the way many fine people have given their heart & soul to iMerit to bring it to where it is. For our customers; the builders, the frontier labs, the enterprises – we are so grateful for being a part of your innovation journey. We look forward to continuing our work & finding new ways For myself, the journeys of our teammates & their role in the global technology revolution called AI, have been heartwarming & rewarding. You achieved your mission & have improved the lives of tens of thousands of people while also helping many hundreds of clients solve some of the most challenging problems in technology. If it was just one or the other it would have been impressive, right? Doing both is the joy of iMerit. I am very thankful to all of you (employees, investors, families) for your hard work, fearless ambition, & team spirit. I ask you all to join me to continue to take this mission forward. Thanks also to the awesome team at Avendus who worked with us on this amazing deal. As we work with our EXL teammates to become one family, I look forward that together we will shape EXL & iMerit into a top AI brand & outperform in the market. Full Release: : https://lnkd.in/gM2CDKQJ #iMerit #EXL Rohit Kapoor
Experience & Education
Other similar profiles
Explore more posts
-
Aviel Ginzburg
While there has never been a more exciting time to be a founder building dev tooling or next-gen infra, it has also never been less investable at seed/pre-seed. I'm either really missing something or a lot of my peers are lost. As someone who has not just written, but also SHIPPED, about 75k lines of code in the past 6 months I can tell you that the evolution of how to build products has changed as much in the past year as it did in the entirety of 2007-2017. The complete rise and fail of frameworks, platforms, methodologies, etc... paved over and forgotten... that is of course except for the 1 company that gets a 1000x return from a wildly overvalued hyper-scaler or drunken growth stage investor obsessed with compounding at scale. Imagine a world where any seed investor in trends like Openstack, Hadoop, PaaS, etc all took a full loss on their investment. That's what we're looking at right now. I personally know of over a dozen well-funded seed-stage companies building in these spaces, with years of runway, scrambling to get acquired for a return of capital + several million personally while they're still relevant. If you're not seeing this unfold in front of you, you either aren't paying attention or you're satisfied playing the lottery instead of investing.
11 Comments -
Chris Kong
The Shenzhen X thread below got me thinking about something I’ve been underweighting. We measure manufacturing success in ROIC and margin expansion. But software taught me to look at something else first: learning velocity. How fast does the system improve per dollar spent? In hardware, that means backing companies that sit inside dense supply chains, not isolated labs. It means pushing for AI- and data-first workflows so iteration cycles are captured and compounded instead of lost. And it probably means influencing policy toward smaller, connected manufacturing hubs rather than betting everything on a few giant centralized nodes. If learning velocity is infrastructure, capital shouldn’t just fund assets. It should increase the rate at which knowledge moves. That’s a different optimization target. https://lnkd.in/gwVSKvbm
-
Norman Volsky🎙️ 🏥 📉
Premiering now on the Digital Health Heavyweights Podcast I have a conversation with the co founder and CEO of Bonfire Analytics Vinay Nagaraj! One great moment was when Vinay says that while business moves at the speed of trust, so does sales. I could not agree more. It requires a huge level of authenticity, being able to connect with a person, as a person. Building deeper relationships. There is a lot to learn on this weeks episode. Vinay shares his journey from biomedical engineering to healthcare sales, discussing the challenges and successes he faced while scaling revenue at Roundtrip. He emphasizes the importance of authentic relationships in sales, the role of data in healthcare, and the key traits he looks for in salespeople. Vinay also discusses the inspiration behind Bonfire Analytics, its mission to accelerate healthcare technology adoption, and the common mistakes healthcare companies make in sales. He provides insights into market opportunities, the importance of making data actionable, and offers advice for founders seeking funding. I always love asking my guests the 3 things they look for when hiring for sales roles, Vinay shares three traits he likes to seek out: 1. Hunger to LEARN, someone always focusing on learning, growing, improving themselves personally, and professionally. 2. Taking Initiative- startups are all the hats, building the plane while you're flying it. So if you see a process that needs fixing, fix it. 3. Strong Communicators- communications that is clear, direct, and easy to understand. I would add radical transparency to this list, but I did only give him 3 traits to distill. Check in to learn if you're a vitamin, or a pain killer, emerging markets, and of course all about Bonfire's impact, and goals, and more! Check out our key takeaways: 🎯 Salespeople should focus on understanding customer pain points to be effective. 📊 Bonfire Analytics aims to provide actionable insights to healthcare companies to improve sales efficiency. 🌍 Vinay's global upbringing fostered adaptability and resilience, essential traits for a startup founder. 🏥 Healthcare companies often make the mistake of relying too heavily on large health systems for sales. 🔑 Data must be actionable; Bonfire helps clients leverage data effectively for their sales strategies. Check it all out here: https://lnkd.in/dnvucCrF
1 Comment
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content