Yacine Rezgui

Recently, I had the opportunity to shadow customer-facing teams at Doctolib, and it was eye-opening. Here’s what I learned that truly resonated with me: 1. Solving Issues on the Fly I was able to witness firsthand how the Support team handles frustrated customers with a solution-oriented mindset. Given that many of our users are healthcare professionals calling in during consultations, speed is crucial. My technical knowledge came in handy when a customer asked about a feature I knew was in development, and it felt great to provide immediate support! 2. Unique Insights into Real-World Usage Seeing our platform in action with actual users was a completely different experience compared to what I’m used to behind a screen. I observed how users interacted with our solutions and pinpointed areas for improvement from both an engineering and UI perspective. I found myself imagining new solutions and enhancements that could elevate their experience. 3. Bringing Us Closer to Our Users At Doctolib, shadowing customer-facing teams is a practice all employees engage in, regardless of their role. The goal is to better understand how our users navigate their day-to-day operations, the challenges they face, and how we can work together to solve them. It’s a simple yet incredibly effective way to stay connected to the real-world impact of our work and move one step closer to building the healthcare we all dream of. #shadowing #doctolib #userfirst #customercentric

59

5 Kommentare

Super excited to share that we just launched 𝗥𝗲𝗿𝗮𝗻𝗸𝗲𝗿 for 𝗩𝗲𝗰𝘁𝗼𝗿 𝗦𝗲𝗮𝗿𝗰𝗵 — making RAG agents even smarter! 🚀 RAG adoption continues to grow, and what really matters is end-to-end quality. But simply giving an LLM more documents isn’t always the answer. Even though large-context models exist, they can still fall short, since quality isn’t guaranteed just by adding context — something I wrote about earlier (https://lnkd.in/g7jJ3mfr). What makes the difference is better ranking that brings the most relevant context to the top when it counts. Building on that, in our paper — 𝗗𝗿𝗼𝘄𝗻𝗶𝗻𝗴 𝗶𝗻 𝗗𝗼𝗰𝘂𝗺𝗲𝗻𝘁𝘀: 𝗖𝗼𝗻𝘀𝗲𝗾𝘂𝗲𝗻𝗰𝗲𝘀 𝗼𝗳 𝗦𝗰𝗮𝗹𝗶𝗻𝗴 𝗥𝗲𝗿𝗮𝗻𝗸𝗲𝗿 𝗜𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 — we found that adding more documents boosts quality at first, but beyond a certain point diminishing returns kick in and quality can even drop because of “phantom hits” (irrelevant passages scored too highly). In other words, quality beats quantity — you want to prioritize the most relevant passages, not just more context. Link in the comments. That’s exactly where 𝗩𝗲𝗰𝘁𝗼𝗿 𝗦𝗲𝗮𝗿𝗰𝗵 𝗥𝗲𝗿𝗮𝗻𝗸𝗲𝗿 comes in. By adding just one parameter to your Vector Search query, you unlock a big quality boost: 𝟴𝟵% of the time the correct answer shows up in the top‑10, a +𝟭𝟱 𝗽𝗼𝗶𝗻𝘁 lift over baseline (74%) and +𝟭𝟬 𝗽𝗼𝗶𝗻𝘁𝘀 over leading cloud alternatives. Really proud of this A+ team that not only built the reranker and retriever at scale, but also co‑authored the paper together with our amazing Databricks Mosaic Research team. Couldn’t ask for a sharper crew! 💪 You can be part of this team too — 𝘄𝗲’𝗿𝗲 𝗵𝗶𝗿𝗶𝗻𝗴 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀 𝗮𝗻𝗱 𝗺𝗮𝗻𝗮𝗴𝗲𝗿𝘀 𝗮𝘁 𝗮𝗹𝗹 𝗹𝗲𝘃𝗲𝗹𝘀 to work on AI products, billion‑scale retrieval systems, and high‑load serving infra. If that excites you, let’s talk! 👤 Senior Engineering Manager https://lnkd.in/g-Nvy64X 👤 Senior Staff Software Engineer https://lnkd.in/g8Yhwx2t 👤 Staff Software Engineer https://lnkd.in/gpHx-nrR 👤 Senior Software Engineer https://lnkd.in/gzFTwnTc

546

18 Kommentare

Future of the User Interface is invisible. This is what I think, the natural langugae will become the norm to interact with any software. You want to book an air ticket? Just chat with an agent on whatsapp, teams, slack, telegram etc and your tickets will be booked. You want to order food online, just chat with an agent, give your preference, and it will place an order for you. The move from clicking buttons to sending messages is the ultimate leap in user experience. ​The next billion-dollar app won't have a beautiful UI. It will have a brilliant, conversation-driven agent.

10

2 Kommentare

Here's exactly how to convert a vibe coded prototype to a real SaaS app in 40 min. In my new episode with Colin (taught 8,500 students AI coding), he shows: ✅ 6 components every real SaaS app needs ✅ Live demo: Converting my AI photo prototype to a real SaaS app with Claude Code ✅ The security flaw in vibe coded apps that most people don't know about 😬 Some quotes from Colin: "A prototype is just a pretty picture. A real SaaS app needs client, server, database, and third-party integrations." "AI makes better decisions if you start with a SaaS template because it's looking at existing patterns instead of building things from scratch." "Many vibe coding tools are insecure. Any user can access any information in your Supabase database if you don't configure it properly." Colin is the real deal for practical AI coding. 📌 Watch now: https://lnkd.in/gUaUMrgU Also available on: Spotify: https://lnkd.in/giN3RmAd Apple: https://lnkd.in/grhhYVyh Newsletter: https://lnkd.in/gSYR9kNw

124

5 Kommentare

Scaling AI Adoption: From 30 Engineers to Tech-Wide Transformation - Lessons from Doctolib's Journey If you're attending The Java Community - Devoxx on October 8th, be sure to catch the presentation by Julien Tanay and Thomas Bentkowski. They'll share Doctolib's journey of scaling AI coding assistants from a 30-person pilot to over 600 engineers. They'll discuss our experiments, including both our failures and successes, to help developers embrace coding assistants. Don't forget to add the session in your favorites ❤️ to help Stephan Janssen with the scheduling. https://lnkd.in/eaGsWDaf

38

3 Kommentare

Why do I write here? For me, writing is a stress test for my ideas. When an idea lives only in my head, it feels brilliant. Clean. Fully formed. But the moment I try to put it on paper, the illusion breaks. I see the idea for what it actually is: incomplete, simple, or embarrassingly vague. Externalizing thinking exposes the gaps, and illuminates the solution path. I didn’t need a psychological study to learn this, I learned it by confronting my own thoughts on paper. My internal rule is simple: If I can’t explain it clearly enough to convince myself, I have zero chance of convincing you. But here’s the truth about my private writing: It’s a mess. If you saw my notes, you’d find scattered words, bilingual phrases, and arguments that only make sense if I’m standing next to you explaining the context. It’s just a raw dump of my brain, a serialization without "form". That is why I post on LinkedIn. Writing in public acts as a forcing function. It forces me to do two things I never do in private: 1. Clarify the idea until it stands entirely on its own. 2. Respect the reader’s time by focusing more on "form", not just content. Beyond that selfish motivation, there’s also a belief I hold: We all carry non-obvious, hard-earned experiences. Most of that knowledge stays invisible, and so many fascinating stories are lost because we never take the time to share them. If your brain works like mine, I encourage you to give writing on LinkedIn a shot. If even one person finds value in what you write, the effort is justified. And if not? At least you walk away with a clearer mind than you started with.

71

4 Kommentare