About
Data and AI leader building AI-native, privacy-safe data foundations that power GenAI…
Articles by Srilatha
-
My First Month at Facebook!
My First Month at Facebook!
It’s been a month since I joined Facebook, and I’ve been asked many times about “What it’s like to work at Facebook”. I…
44 Comments
Activity
5K followers
-
Srilatha Kothur shared this38 AI B2B startups in one room at the a16z speedrun AI Faire last Friday. Robots, voice AI, developer tools, sales automation. But the companies I kept coming back to were the ones nobody would put on a hype reel. Infragrid records how experts actually work across browsers and applications and turns that into deterministic, repeatable workflows with built-in approvals and auditability. Financial reviews, compliance evaluations, fraud investigations. Phyvant captures the judgment, business rules, and institutional knowledge that lives in your best people's heads and turns it into a system the organization owns. Tax advisory, M&A due diligence, compliance audits. Their pitch: "Buy the system, not the model." The model is common property. Your judgment is not. Truli automates FDA and FTC regulatory compliance for food, supplement, and CPG brands. Live parsing against the actual Code of Federal Regulations, snack bags on the table next to a CFR browser catching label issues most brands wouldn't find until enforcement did. The thread across all three: AI going after work that's complex, regulated, and high-consequence. Scaling it, and making it consistently right in the places where wrong has real cost. The best companies in that room weren't chasing what's obvious. They were solving what's hard. #EnterpriseAI
-
Srilatha Kothur shared thisiCloud is hiring in Shanghai. High-impact roles at massive scale: Data Engineering, Data Science, ML Optimization and SRE, all supporting one of the largest cloud services in the world. If you're in China (or your network is), take a look. Links to each role are in the original post below. Know someone great? Tag them or pass it along.Srilatha Kothur shared thisiCloud is hiring in Shanghai, China! Roles span iCloud SRE, Data Science, Data Engineering, and ML Optimization. For those who have networks that include China, please pass along! I am not the hiring manager so please click the links to get more details and/or apply! iCloud SRE: https://lnkd.in/gvJkiRdQ iCloud Data Engineer: https://lnkd.in/gcY87hPP iCloud Data Scientist: https://lnkd.in/gCmgxFam iCloud ML Optimization Engineer: https://lnkd.in/gWSMuPpaSenior Site Reliability Engineer - ASE / iCloud - 招贤纳才 (中国)Senior Site Reliability Engineer - ASE / iCloud - 招贤纳才 (中国)
-
Srilatha Kothur posted thisNothing meaningful gets built alone. It gets built by people who saw the problem, saw the conviction behind it, and decided: “I’m in.” No one gives up a weekend to optimize an ad click. But the moment the work actually matters, everything changes. People carve out time to offer their expertise. Busy executives answer cold emails. Strangers become advisors, then champions. Generosity follows meaning. The world holds an enormous reserve of goodwill, waiting for a worthy reason to come alive. What calls it forward is someone willing to look at a problem everyone tolerates and say: not anymore. People don't lend their weekends to ideas. They lend them to conviction, fresh thinking, and the courage to bet on both. That's how communities form around meaningful work. That's the untold story of this moment in tech. The headlines are about AI, competition, and disruption. The truth on the ground: more people are creating than at any time in memory. They're doing it together, handing each other playbooks that would have been trade secrets a decade ago. Behind every breakthrough is a quiet chorus of people who said: "I'm in." So this week, two things: Thank someone who helped you when they didn't have to. And if your expertise is sitting idle, lend it to something meaningful that matters to you. You might be the reason it works. #TechForGood #Leadership #PayItForward
-
Srilatha Kothur shared thisThe most useful thing AI did in our house this World Cup was show my kids exactly where it stops. My boys built a bracket board and painted every flag by hand. Before each match they lock in a prediction, then ask an AI for its call, and track who is beating whom. Every dot in that middle eventually gets filled by reality. A month in, two opinionated kids are holding their own against the model. The moments they will actually remember, nobody predicted. Messi breaking a scoring record. A streamer's chaos anthem ending up on the official FIFA album. Sixty thousand people from every corner of the planet losing their minds in unison at Levi's Stadium. None of it was in anyone's forecast. That is not a flaw in the tournament. It is the point. Models are becoming extraordinary at finding patterns from what has already happened. But human intelligence has to reason about a world where many futures are possible at once. Watch any match and you see the shape of it. A player reads the ball, the defender, the space, and the dozen things that could happen next. That is the difference I keep noticing on our couch. My kids hold a dozen futures at once, and they know it. They are not just asking, "What will happen?" They are asking, "What could happen?" And that matters beyond a soccer field. Then there is the money, and the scale is staggering. Biggest tournament in history. 48 teams, 104 matches, 16 cities. Tens of billions in GDP projected. Every major event brings the same question leaders wrestle with: what value can actually be measured? Some things show up in revenue numbers and economic reports. Some things show up years later: in a child's memory, a city's identity, or a community coming together. Nobody argues about the roar of sixty thousand people, or the memory my kids will carry for life. The surest value in that stadium is the part no one can put a number on. It never reaches a balance sheet. It belongs to the people who were there. The skill was never predicting everything. It is knowing what is worth predicting, and what is worth simply experiencing. My kids figured that out on a board they painted themselves. #FIFAWorldCup #DecisionIntelligence #AI #Leadership
-
Srilatha Kothur shared thisFor every hour of real work, there's another hour spent proving the hour happened. The status update. The deck nobody opened. The report that took a week to write and ninety seconds to skim. I call it the "proof tax," the time we spend proving we did the thing instead of doing it. It shows up everywhere: in code, in classrooms, in clinics, and in corporate budgets. And it's heaviest exactly where the stakes are highest: → The nurse who spends more of the shift charting the care than giving it. → The teacher documenting that learning happened instead of making it happen. → The analyst rebuilding the same budget report for the fifth audience instead of helping decide what's in it. Here's the honest part: some proof isn't waste, it's the work. The trail that lets the next person pick up safely. The accountability that earns trust. So it was never proof vs. work. The real line is this: does the proof change what happens next, or does it just record that it happened? One kind compounds. The other only costs. In his book Nexus, Yuval Noah Harari notes that human networks rely on bureaucracy to create order, but warns that these information systems easily become self-serving, prioritizing their own internal records over external reality. Today, automation and AI are making it effortless to generate reports. But that doesn't shrink the pile. It multiplies it. The constraint was never how fast we could generate the record. It was whether we ever stopped to ask which records were worth generating. None of this came from bad intentions. We built every layer for a reason. But we forgot to ask what it was costing us. We have to start building systems that give the hour back to the people doing the work. Systems that automate the baseline tracking so human beings can focus on what actually changes what happens next. Lately, I've been building toward a fix for this ratio. What's your current ratio of mission to proof? And what would you build if it flipped? #ProductivityParadox
-
Srilatha Kothur posted thisLast week at an AI conference, I sat with a group of women founders building in healthcare, climate, and education. The kind of people you walk away from thinking, okay, the future is in good hands. The conversation turned to funding, and the table split in two. Half hadn't realized how much was there. Small Business Innovation Research grants, Small Business Technology Transfer programs, National Science Foundation awards, dedicated set-asides for women-owned businesses. Real programs, significant dollars, built specifically for the kind of work they were doing. The other half had tried and described a different challenge: navigating the process itself. One founder summed it up perfectly: "VC says no in two weeks. A grant says maybe in nine months." What stayed with me wasn't a funding gap. It was an access gap. The women at that table weren't lacking ambition, expertise, or strong ideas. They were building serious companies. The challenge wasn't finding opportunities. It was figuring out how to navigate them. It reinforced a pattern I keep seeing across industries: sometimes the biggest bottleneck isn't technology. It's friction. The opportunities may already exist. The hard part is making them reachable. If you’ve pursued grants or non-dilutive funding, what was harder: finding the right opportunities or navigating the process itself? #WomenInTech #WomenFounders #WomenOwnedBusiness
-
Srilatha Kothur posted thisThe AI coding revolution has solved one problem: speed. It has not solved the harder one: accountability. We can now go from idea to working product in a weekend. But most AI tools optimize for creation, not operation. Cursor won't ask about your data retention policy. Claude Code won't stop you from sending PII into a prompt. Replit won't flag that you passed an entire database row to a model when two fields were enough. The default path is: send more context, get a better answer, move faster. That works until the stakes change. And in the domains where AI is headed next, the blast radius of a bad decision isn't a wrong recommendation. It's someone's health, someone's livelihood, or how public money gets spent. We will look back at this era the way we look back at early internet before HTTPS. Everything worked. Nothing was secure. And we only fixed it after the consequences forced us to. AI doesn't have to repeat that cycle. But it will unless we treat accountability not as an afterthought bolted on after launch, but as runtime infrastructure shipped with the product from day one. Context boundaries, traceability, policy checks, human escalation paths, and fallback modes built in, not added later. The next frontier is not faster AI development. It's AI that can account for itself while it runs. Systems that know what data they used, which policy they applied, which model made the call, and whether a human should have been in the loop before the decision reached someone. The AI that gets adopted in high-stakes domains won't be the one that ships fastest. It will be the one people can trust after it ships. #AI #ResponsibleAI #TrustworthyAI #AIAccountability
-
Srilatha Kothur posted thisLots of layoff posts in my feed today. I want to share something different. Over the past year, I've met women who were impacted by earlier rounds of tech layoffs and went on to build companies solving real problems for people like them. Problems they'd been thinking about for years but never had the space to pursue. The pattern was the same every time. The layoff forced a clarity they wouldn't have found on their own. It created space to ask what they actually cared about and go build for it. One told me, "I spent years optimizing systems that were already world-class. Now I'm building for people who have nothing." They're building with a depth of conviction and operational expertise that only comes from years at the highest level of tech, applied to problems they deeply care about. To everyone impacted right now: the grief is real. But when the fog clears, know that some of the strongest founders I've met didn't choose this path. The path found them after the ground shifted. We're about to see a new wave of builders creating things that genuinely matter.
-
Srilatha Kothur posted thisThere's no shortage of big AI conversations right now. Model capabilities, enterprise transformation, agentic systems, the race to AGI. All of it matters. But the AI use case I keep coming back to is the one I rarely see discussed: using AI to deeply understand problem spaces previously inaccessible from the outside. After 18 years building data products and infrastructure at Apple, Meta and Netflix, I started using AI on my own time to go deep on domains I've never worked in. Studying what data exists publicly, what real problems people face, how decisions get made, and where AI could actually matter. It started with my own life. I'm a woman navigating hormonal volatility with data everywhere. Wearable, health apps, lab results, provider notes. None of it connects. Friends going through PCOS, fertility journeys are doing the same. Cobbling insights from five apps, bringing Oura screenshots to their OB-GYN, Googling lab ranges at midnight. Women's health may be the most data-rich yet insight-poor domain in healthcare. That taught me a pattern: the hardest data problems aren't where data is missing. They're where data exists everywhere but nothing connects it. I saw it everywhere. Our renovation took years. Permits on paper, bids in spreadsheets. Even the largest construction firms stitch together fragmented systems in ways that would horrify anyone in tech. I pulled the thread into public infrastructure. How does federal funding flow from DC to a contractor in your neighborhood? How do people managing public money prove compliance when data lives in ten disconnected systems? Every domain had the same problem: critical workflows on fragmented systems, domain expertise carried in people's heads, almost no intelligent tooling built for how they actually work. I built multiple prototypes. Not to ship products, but because these problems felt too important to just observe. AI gets you to 60-70% fast. Then it hits a wall. The wall is domain expertise. The kind in the heads of people who've spent decades inside these systems. The judgment, the edge cases, the "technically yes but in practice never" knowledge. This journey introduced me to incredible professionals who confirmed these problems are real, and were genuinely curious whether AI could finally address them. Their expertise reshaped how I understood these systems. Some of it works. A lot doesn't, yet. In domains like public finance and healthcare, that gap has real consequences. An audit finding. A misdiagnosis. A community waiting for something that matters. The most important AI work of the next decade won't be in industries with world-class tooling. It'll happen where people navigate enormous complexity with far less. Not replacing their expertise. Augmenting it. Some of these domains have pulled me in deeper than I expected. I'll be sharing more of what I'm learning. If you work inside any of these spaces, I'd love to hear your perspective. #BigIdeas2026 #BuildingwithAI
Srilatha Kothur commented on a post
1mo
Congrats Roi. After working alongside you on Shops at Meta, I'm not surprised at all. It's great to see you channeling that same energy into such an important problem. Rooting for you and the Glow team!!
-
Srilatha Kothur reacted on thisSrilatha Kothur reacted on thisI'm excited to share that I'm joining OpenAI as a Member of Technical Staff, focused on Trust & Safety, reporting to Paul Ellwood, who leads Data Engineering. I start next week. Huge thanks to my friend Gabriel, who kept pushing me to seriously consider OpenAI. A first conversation with Paul turned into a full interview loop, a visit to the office, and ultimately signing the offer. Although I left Meta two months ago intending to take a break, I was drawn in immediately. Coming from a Privacy background, working on Trust & Safety and Integrity at this moment feels deeply meaningful. There is a lot to solve, and the team is moving fast. I’m especially excited to help rethink the Data Engineering craft for a world where agents let us build faster and reshape workflows to be more autonomous and AI-native. Paul and David Sasaki, who leads Analytics, have built a fantastic team with a lot to learn from. Thanks also to my recruiters, Kristen and Omari, for being so diligent and accommodating throughout. I’m excited to start building alongside DE OGs and former Meta colleagues Bryan and Bharat, and to reconnect with Yihong, Mahesh, Flynn, and many others already there. It is time to crank up the intensity again :-)
-
Srilatha Kothur reacted on thisSrilatha Kothur reacted on thisI came to Switzerland to listen. I ended up covering a lot more ground than I expected, both professionally and literally. Over the past several days, as part of my ICMA Tranter-Leong Fellowship, I traveled across the Lake Geneva region, meeting with leaders working at the intersection of AI, government, science, innovation, institutional autonomy, and public trust. The conversations took me from ETH Zurich and Switzerland’s emerging AI ecosystem, to Trust Valley and Innovaud, to the Canton of Vaud, and finally to GESDA, the Geneva Science and Diplomacy Anticipator, where we discussed how this work might contribute to the broader “Road to Geneva” and the global conversation leading toward the “Geneva AI Summit 2027.” A central idea became clearer with every conversation: Public institutions do not simply need access to AI. They need the capability to use it in ways that preserve institutional autonomy, protect data and knowledge, strengthen rather than diminish human competence, and ensure that public decisions remain accountable to the people and institutions authorized to make them. That is a much bigger challenge than adopting another technology. And because sitting still between meetings is not one of my strengths, I also experienced Switzerland from the saddle. I rode roughly 212 miles around Lake Geneva through cities, vineyards, villages, and lakefront roads. For the finale, I rode approximately 220 miles and 9K of climbing to the Col du Grand-Saint-Bernard and the Swiss-Italian border. There were spectacular views, brutal climbs, questionable navigation, flat tires, and memories I will always cherish. But the rides became part of the listening tour. Moving through the country slowly gives you a different appreciation for Switzerland: the geography, the cantons, the infrastructure, the proximity of communities and institutions, and the extraordinary concentration of science, government, diplomacy, and innovation in such a small place. I leave Switzerland with pages of notes, dozens of new relationships, a much sharper research thesis, invitations to continue the conversation in Geneva, and plans already taking shape for an invitation-only roundtable at Pepperdine University’s Château d’Hauteville this November. Most importantly, I leave with more questions than answers, which is exactly what a listening tour should produce. There is much more work ahead. Thank you to ICMA, Pepperdine, and the wonderful people who welcomed me. Hana Disch, Alexandre Meldem, Pascal Marmier (孟思恺), Charlotte MOUNIER, Patrick Barbey, Marie-Laure Gallez, Lennig Pedron, Jeremy Lovey, Jan Kerschgens, Catherine Pugin, Franck Dessoly, Joanna Wiśniewska, Justine Duvernay, Olivier Reynaud, Marc Goodman #ArtificialIntelligence #LocalGovernment #PublicLeadership #DigitalGovernance #PublicTrust #Switzerland #Geneva #GESDA #ETHZurich #Pepperdine #DavenportInstitute #ICMA
-
Srilatha Kothur reacted on thisSrilatha Kothur reacted on thisWe built Wajo AI with the core belief that agents needed to be 1. access controlled, knowledge scoped, 2. accountable i.e. come with their own identity, inbox, phone capabilities and credit cards 3. be proactive, self learning and steerable i.e able to get $h!t done, whether it means by using their computers, calling someone or texting them. Official launch isn't out yet, but the underbelly of the valley is already using these agents. This vision has come to life, and it's kind of insane to see this blow up. Come to our launch happy hour in Wajo HQ :) https://luma.com/hdwnvogh
-
Srilatha Kothur reacted on thisSrilatha Kothur reacted on thisSpent last week up in Sacramento and I'm still kind of pinching myself. I was named a 2026 California Woman Founder of the Year, one of eleven founders honored this year for pushing the boundaries of innovation and growing the state's economy. Huge thank you to First Partner Jennifer Siebel Newsom and Senate President pro Tem Monique Limón for putting real weight behind investing in women-led ideas. I also got to sit down with State Senator Josh Becker (a friend since we were in our 20's and an incredible advocate for Silicon Valley). Here's the thing. Only in California (maybe only in Silicon Valley!) does a 52-year-old woman get to look at menopause and think, I'm going to build a company around this and raise the capital to do it. That's what makes this state extraordinary. My hope is that the innovation doesn't stop here — that the spirit of it, and the funding behind it, keeps spreading to women building big things everywhere.
-
Srilatha Kothur reacted on thisSrilatha Kothur reacted on thisWe're bullish on India. South Park Commons Bangalore was our first bet outside the US. Some of the most ambitious projects of the decade will be built there. Indian dynamism is here to stay.
-
Srilatha Kothur reacted on thisSrilatha Kothur reacted on thisWe've been cooking over at Ping AI in private early access. Here's a sneak peek of our messaging platform for humans and agents. We've just launched voice mode and it's already the only way I want to interact on mobile.
Experience & Education
Recommendations received
8 people have recommended Srilatha
Join now to viewOther similar profiles
Explore more posts
-
VarenyaZ
Start with one narrow AI use case tied to a measurable KPI, and prove it before you scale. Plans fail when teams skip data readiness and regulatory constraints and try to deploy a large, multi‑use‑case solution before a real pilot. Use a three‑step framework: select 1–2 KPI‑driven use cases, verify data quality and US regulatory constraints, and choose a delivery model with a phased pilot and explicit success criteria, while embedding privacy, security and change management from day one and budgeting for ongoing costs. Read the full guide to structure your AI integration program. https://lnkd.in/eZWjcsuw Learn more at https://varenyaz.com/ and contact us at business@varenyaz.com. #AIIntegration #DataGovernance #RegulatoryCompliance #USBusiness #TechLeadership
-
CMEx Management Private Limited
With IBM Instana, on-call engineers get real-time visibility across every layer of their application stack. Color-coded insights make it easy to spot performance deviations, track system health and resolve issues before they impact users. No more alert fatigue. No more guesswork. Just actionable observability that keeps your apps performing at their best. Discover how IBM Instana simplifies complex monitoring for modern teams. https://lnkd.in/eezCgmeH #IBMInstana #Observability #AIOps #DevOps #Monitoring #PerformanceInsights #IncidentManagement #Uptime #CMEx #CMExcellence
-
Nuage
Reddit is signaling a clear message for mid-market companies. At Nuage, we see a common theme emerging: stabilize your foundation before layering on AI. Clean data, solid processes, and strong system adoption are essential first steps. Once these are in place, AI can deliver real, meaningful value.
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content