À propos
👋 I'm Val.
I built my first website at 14, then couldn't choose between business…
I built my first website at 14, then couldn't choose between business…
Activité
18 k abonnés
-
Valentin Huang a partagé ceciWe built a Chrome extension that puts a green or red pill on every LinkedIn profile we open. 🟢 means the company fits our ICP 🔴 means move on. We prospect on LinkedIn, so every new profile came with 3 minutes of tab-opening. Company page, people tab, count the product folks, insights tab, check headcount growth, lose the thread of the conversation. Multiply this by hundreds of profiles a week, that's a part-time job nobody signed up for. So we built an extension that answers these questions in about 3 seconds, on profiles and inside conversations, based on our ICP signals like: - how many people work in product there - headcount growth over the past 12 months We're sharing the whole thing. It ships with example thresholds (3+ product people and growing) so you can see the mechanics. Yours will be different: change 4 lines of config, or drop the folder into Claude Code and describe your ICP in plain English. That's how we built it in the first place. Runs locally on your own session. No server, nothing leaves your machine. Package and setup instructions in the first comment.
-
Valentin Huang a partagé ceciWe're making Harvestr more collaborative. Here's the product thinking behind it: Almost 100% of the product teams I talk to now use Claude. So does part of our own team. When your main competitor is also your teammate, you get very clear, very fast, on where you can differentiate. Here's where we landed: LLM usage is single-player. One person, one context window, one output. The thinking is excellent. The artifact dies in a chat nobody else opens. Product work doesn't run that way. Deciding what to build is a group decision by nature. A PM, an engineer, a CSM and a founder arguing over the same context, each holding different information. We know we'll always lose the race to whoever ships the next frontier model, and we'd rather build on top of it. So we're betting on other things like collaboration around the decision. First piece shipped this week: collaborative discovery docs. → Comment on any passage of a discovery doc, anchored to the exact text → Threads, replies, resolve, mentions → Anchors survive edits and flag when the text has changed → Images, code blocks, click-to-edit Nothing revolutionary as a feature list. Notion and Google Docs have had it for years. What changes is where it sits: on top of customer feedback, product insights and the roadmap, so the argument happens next to the evidence instead of in a Slack thread three days later. A model can tell you what customers are asking for. It can't get your team to agree on what to do about it. That part is ours 💪
-
Valentin Huang a partagé ceciEvery Monday, Claude tells our team everything our AI agent for product teams got wrong last week. Errors, incomplete answers, misunderstandings: every gap it surfaces is our next feature. While building HAI, our AI agent for product teams, we learned fast that an AI agent is only as good as your rhythm of iterating on it. So we instrumented every conversation with Arize AI for full tracing, and handed the weekly analysis to Claude: → reads last week's real conversations, all traced in Arize → clusters what customers asked for → flags where the agent failed — wrong counts, "are you sure?", hallucinations → tracks cost + latency → posts it to Slack, tags the team, with a link to every session behind every claim → and hands us the fixes as ready-to-pick-up tickets We turned it into a Claude routine so it fires on its own, every Monday morning. How are you closing the loop on your AI agents? 👇
-
Valentin Huang a partagé ceciPM: "We run a fully automated eval pipeline across all our AI features." Meanwhile, the pipeline:
-
Valentin Huang a partagé ceciWe doubled the accuracy of our feedback auto-categorization system. Still zero training required from customers 🦾 Customer feedback is messy. Support tickets, call transcripts, Slack threads, sales notes. Every customer's data looks completely different. And we don't stop at theme level. Anyone can bucket feedback into "pricing" or "onboarding". We categorize down to the actual customer problem and the feature it maps to. That granularity is what makes it actionable for product teams. It's also the toughest part to build. The lazy fix is to make customers set it up: define a taxonomy, tag examples, train, maintain it forever. We didn't want that. Harvestr has always been self-serve, so the system had to be self-learning. So we kept iterating on what became a proprietary pipeline built over years, tuned on real production data, until accuracy doubled 📈 Very proud of what the team shipped. This is a genuinely hard problem to solve, and more improvements are already in the pipeline. Try it on your own feedback today 👉 https://lnkd.in/eWhyWMDK
-
Valentin Huang a partagé ceciI connected Metabase and PostHog to the same Claude session and it just changed how I do product discovery. The old workflow: 🥱 Run SQL in Metabase, spot a weird usage pattern 🥱 Open PostHog, try to reconstruct which users and which timeframe 🥱 Watch 15 sessions hoping some match the pattern 🥱 Lose half the context in between The new one: 😍 Ask Claude to query Metabase on a discovery topic. Quantitative signal comes back in seconds 😍 Spot a behavior worth digging into 😍 In the same session, ask PostHog to find the exact user sessions behind those data points 😍 Watch exactly the sessions that matter, not a random sample The quant analysis produces the filter for the qual analysis. Numbers tell you what's happening, sessions tell you why. Having both in one conversation is 💯
-
Valentin Huang a partagé ceciPM: "I ship code to production every day." What every day looks like:
-
Valentin Huang a partagé ceciPMs can write code with AI. But can they ship it without clogging the review queue, breaking the build, or spending hours they don't have learning git workflows? One-off changes are easy. Doing it consistently, as a non-technical PM, on a production app, is a different story. So we built a Claude Code skill that handles the process: 1. Focused on front-end code. It's where PMs can ship the most value with the least risk. 2. Deterministic guardrails. Every change gets classified before a line is written. Anything touching architecture, data access, or complex layout gets stopped and handed to a dev. 3. Scoped to where I can ship value. Copy, config, i18n, isolated tweaks. Nothing that requires you to guess. 4. Full integration with our existing pipeline. Lint, build, git hooks, MR, Slack ping, Shortcut link. I don't run a parallel workflow. 5. Tech review on every MR, no exceptions. This helped us build a process that both product and engineering trust. I use it to ship code to production every week. We turned it into a template you can adapt to your own stack, CI/CD, and review process. Comment "Ship" and I'll share it with you 🛠️
-
Valentin Huang a aimé ceciValentin Huang a aimé ceciChez METRO France en 2026, on ne vend plus de vin. Non, on aide nos clients à en vendre. On leur apporte du service, des solutions. Car si notre métier a longtemps consisté à parler d'AOC, de terroirs et de domaines, nos clients ont surtout besoin de réponses concrètes : comment construire leur carte, piloter leur coût portion, choisir les bons vins pour leur concept et développer leur rentabilité. C'est tout l'esprit de notre nouvelle Collection Vins 2026 🍷, qui arrive aujourd'hui dans nos halles et en ligne : des exemples de cartes des vins pensées par typologie d'établissement, des conseils de sommeliers pour favoriser les rotations et valoriser l'offre, mais aussi un coût de revient au verre affiché pour chaque bouteille. Une approche que nos cavistes en halles portent depuis longtemps au quotidien, au plus près des restaurateurs. Parce que le meilleur service que nous pouvons rendre à nos clients, c'est de leur apporter des solutions utiles et immédiatement actionnables. Ce changement de logique est une vraie aventure collective dont je suis particulièrement fière. Bravo à toutes les équipes achats, pub & marketing, supply, merchandising, communication et commerce qui y ont contribué. Samirah HEINIMANN Quentin Bodet Jean-Luc Charroud irene elbaz Christine Chan Caroline Marie Inna RIBETON Clara ROCHE Fabrice Cavau Marion RODRIGUES Samuel Deroo Didier Henaux Léopold Dufour Lucile Morison Christophe De Baere Sebastien Lacresse Loren Benmussa Emma Munier Damien Melot Tiphaine PANTIN Aymeric LEPREUX Merci également à tous nos fournisseurs qui viennent faire déguster, animer et vivre nos produits en halles. Dans un contexte plus difficile que jamais pour le secteur du vin, cette collaboration est précieuse ! Et maintenant, place aux ventes🍷 Catalogue en commentaire 👇
-
Valentin Huang a aimé ceciA huge milestone years in the making. Proud to be part of this journey and incredibly proud of the entire Litmus team! Thank you to our customers and partners who continue to push us forward. Major win unlocked. More to come! 🚀🔥Valentin Huang a aimé ceciWe’re proud to announce we’ve been named a Leader in the 2026 Gartner® Magic Quadrant™ for Global Industrial AIoT Platforms. Litmus provides a unified Industrial Data Foundation across all sites: one common layer to connect, contextualize, govern, and deliver factory data to enterprise AI, without rebuilding the architecture for each site. That’s the whole idea. Not a platform for one plant, but a foundation for all of them. It’s what lets manufacturers take an AI initiative that works, run it across the network, and reliably receive consistent data, governed the same way, at every site. We believe this recognition reflects the growing importance of that foundation as manufacturers move from AI pilots to enterprise-scale deployment. To every customer who has built on Litmus and every team member who has shipped it—thank you. This is yours. See what it means to us for scaling Industrial AI → https://hubs.li/Q04x_0Bp0 Download the report here: https://hubs.li/Q04x_0wg0
-
Valentin Huang a aimé ceciValentin Huang a aimé ceciWe're hiring a founding mobile engineer at Nory - ideally based in Barcelona! 🇪🇸 Not "join the mobile team." Build it. You'd work directly with me and the awesome Pablo Deeleman & Igor Kochajkiewicz on this, 0 to 1! & trust me, we have some very cool stuff in the pipeline 🤫 😉 🦄 If you want founding-engineer energy, at a company where mobile actually matters (restaurants live and die by what happens on the shift), this is the one! 📱 Apply here: https://lnkd.in/e7UZFB6H
-
Valentin Huang a aimé ceciValentin Huang a aimé ceciSales people in 1995: whip out the phone book, thumb through the pages until you find the right contact, smile, dial. Total time to book 5 meetings: days. Sales people in 2026: whip out one tool for data, another for enrichment, still another for sending, and something else for dialing. Total time to book 5 meetings: STILL DAYS. At Apollo.io, we truly believe there's a better way. That's why we built THE AI GTM system for modern revenue teams. Travel back in time and learn more here: https://lnkd.in/ew2mAiJH Time to upgrade.
-
Valentin Huang a aimé ceciIf you are interested in this role, don’t hesitate and reach it to me ����Valentin Huang a aimé ceciMost HRBP roles come with a playbook. This one comes with a blank page. Since March, we've been building a People team at Kolsquare | B Corp™ I'm really proud of. This is the last piece of the puzzle. We're hiring the first HR Business Partner for Kolsquare | B Corp™ and helloCash / mRaP GmbH, two team.blue companies. The role doesn't exist yet in either one. You'll be the person who shapes it. What's on the table: 🤝 Being the go-to partner for around 35 managers and our works council 📈 Shaping our performance cycle and team development paths 🧩 Leading the people side of a real post-acquisition integration (Kolsquare | B Corp™ x Storyclash) 🚀 Direct access to leadership, with the stability of a profitable group behind you If you're an experienced HRBP in Austria who'd rather build than maintain, check the job below. Remote first, based in Austria. German and English (French is a plus). #hiring #HRBP #Austria #remote
-
Valentin Huang a aimé ceciYour users are definitely using AI. And most organizations don't quantify the value OR cost of their data - if it's going into a public-facing platform, assume it's being retained, analyzed, or worse: being used to train new models. We'll talk through the risks - and also - the potential solutions. Come see me!Valentin Huang a aimé ceciThe conversation around campus AI is shifting from whether institutions should respond to how they can respond strategically. 🎤 On September 22, Cloudforce speakers Brian Dreyer and Joey Poole will join Carahsoft to share their perspective on Shadow AI and the path toward a more governed, campus-wide approach. Attendees will learn: ▪️What Shadow AI can reveal about unmet demand across campus ▪️How to evaluate AI use through the lenses of demand, data, experience, and governance ▪️What institutions need to support AI adoption with greater visibility and coordination 🤖 If these questions are part of your institution’s AI strategy, join us for From Shadow AI to Strategic Advantage. Save your seat: https://lnkd.in/eY8QqAcQ
-
Valentin Huang a aimé ceciReally exciting to see how far we have come over the years.Valentin Huang a aimé ceciSkimmer has launched Working Capital through Stripe — a new financing offering that gives eligible pool and spa service businesses access to fast, flexible financing they can use to invest in new trucks, expand into additional routes, or manage cash flow during slower months. Eligible businesses are automatically prequalified for a personalized financing offer, with funds typically deposited in as little as one to two business days if approved. Learn more: 👇 https://lnkd.in/gxwHC9WN Stripe Capital offers financing types that include loans and merchant cash advances. All financing applications are subject to review prior to approval. In the US, Stripe Capital loans are issued by Celtic Bank, and YouLend provides Stripe Capital merchant cash advances.
-
Valentin Huang a aimé ceciI've spent over a decade working in and watching the tech ecosystem in St. John's emerge, evolve, and establish itself as a global leader. Could not be more proud of how far we've come, and how much more we have in the tank.Valentin Huang a aimé ceciBehind the fog, St. John's is a hidden technology gem. Verafin exited for $2.75B in 2020. CoLab and Spellbook are on track to be unicorns in short order and at the top of the leaderboard in two of the most important AI races: engineering & legal. Kraken Robotics is trading at north of $2B market cap and is a global leader in subsea intelligence. There is a thriving student ecosystem led by the Student Design Hub - Memorial Medtech won nationally this year and Eastern Edge Robotics won internationally. How is this possible for such a small place? The people. Beyond being some of the friendliest and fun folks on the planet there are some of the smartest people you'll find anywhere, here. The best people are ultra gritty and have a chip on their shoulder to prove it is possible to win from this rock in the middle of the atlantic ocean. That's a dangerous combination that makes anything possible - including punching well above our weight class for a small place. Thank you to the The Globe and Mail, Sean Silcoff and Alexandra Posadzki for this exceptional article. Proud to be a Newfoundlander. LFG! 🤘 https://lnkd.in/g_2FJWVYHow St. John’s has become an unlikely hub for tech startupsHow St. John’s has become an unlikely hub for tech startups
-
Valentin Huang a aimé ceciValentin Huang a aimé ceciIt’s been a few days since Ventrata Open, and I’ve finally had some time to take it all in. I still can’t quite believe someone let me do this? Speaking in a real cinema, full of 300-something people, about something I care deeply about was a wild experience. I got to talk about our web products and how we’re planning to make life easier for our clients and their users in the near future. Then the next day, I got to sit down with smaller groups of around 30 people for workshops on the CMS and its visual editor (my baby). Those conversations were a highlight for me. Showing people what it can do, hearing their feedback (and feature requests), and finding out how they’d use it and what it could replace for them was awesome. As much as I love being remote-first, there’s really nothing like talking face to face with people who already use your product, or might use it next. After months of preparation from everyone involved, hearing that people enjoyed the sessions and got something useful out of them felt really good. I got a ton out of them too. Really grateful to the team for their support and trust, to everyone who put so much into making it all run smoothly, and to all the clients and partners who joined us. Happy I got to be part of it.
Expérience et formation
Langues
-
French
Bilingue ou langue natale
-
English
Capacité professionnelle générale
-
German
Compétence professionnelle limitée