À propos
I specialize in designing, building, and optimizing large-scale infrastructure, with…
Activité
539 abonnés
-
Amine Kherbouche a partagé ceci💪Amine Kherbouche a partagé ceci215 !! Et oui 215 collaborateurs, très exactement un an après notre pivot, nous avons triplé de taille et onboardé 145 ingénieurs et développeurs en faisant la promotion de notre culture d'entreprise, de notre environnement technique. Une formidable aventure humaine, avec ses hauts, ses bas, ses trahisons, tout à construire et à faire, mais surtout une opportunité incroyable de s'entourer de talent de tous les horizons et de toutes les nationalités. J'aime mes gars !!! <3
Amine Kherbouche a commenté un post
5 ans
Congrats :*
-
Amine Kherbouche a aimé ceciAmine Kherbouche a aimé ceciPresented Capsule at the Claude meetup organised by SLOAI 🎉 Capsule is our internal Agentic Developer Environment at Teads — a platform that gives Claude Code its own isolated, secure sandbox to work autonomously on your repos, without running with your full privileges. Thanks for the warm welcome and the lively Q&A! 🙌 #ClaudeCode #AgenticAI #DeveloperTools #Teads
-
Amine Kherbouche a aimé ceciAmine Kherbouche a aimé ceciI recently spent quite a bit of time chasing a surprisingly tricky issue in our billing pipeline at Gorgias. Everything worked perfectly in real time. Then we replayed historical events. What followed was a deep dive into Apache Flink event-time processing, watermarks, repartitioning, and some behaviors that were much harder to reason about than the documentation would suggest. I wrote up the investigation, the dead ends, the hypotheses, and the fixes we eventually implemented. If you're working with Kafka, Flink, event sourcing, or large-scale historical reprocessing, you might find it useful: https://lnkd.in/ee5WQGJR Thanks to the teammates who helped review the article and challenge some of the assumptions along the way.When Event Time Meets Reality: Lessons from Building Billing on Apache FlinkWhen Event Time Meets Reality: Lessons from Building Billing on Apache Flink
-
Amine Kherbouche a aimé ceciAmine Kherbouche a aimé ceciEveryone wants a full remote or hybrid role. But very few candidates truly understand what it means to interview for one. Remote work is not just a benefit. It is a trust-based operating system. And trust is not given because your CV looks strong. Trust is built through the way you communicate, prepare, follow up, clarify, and create confidence in others. I have seen too many strong candidates lose momentum in a process not because they lacked technical skills, but because they didn’t show the behaviours required to succeed in a remote-first environment. They came unprepared. They had not done the homework. They struggled to explain clearly why the company, why the role, why now. They gave long, unstructured answers in a 30-minute interview. They communicated in a way that felt too casual, too passive, or not sharp enough for the level of autonomy expected. In a remote or hybrid company, your communication is part of your performance. When people cannot observe how you work every day, they rely on signals. How clearly do you explain your thinking? How concise are you when time is limited? How well do you structure ambiguity? How proactive are you in creating alignment? How quickly do you follow up? How much ownership do you show without being chased? How much confidence do you create in the people evaluating you? This is especially true in high-performance environments. The best remote candidates understand that being good at the job is only one part of the equation. You also need to show that you can operate with autonomy, communicate with precision, build trust asynchronously, and reduce uncertainty for the people around you. Being sharp, concise, and structured is not a “nice to have.” It is a must-have. Being proactive is not a bonus. It is a trust-building mechanism. Preparation is not about pleasing the interviewer. It is about showing that you respect the opportunity, the company, and the time of everyone involved. Because in remote-first teams, the cost of poor communication compounds fast. A vague update creates confusion. A lack of follow-up creates doubt. A passive attitude creates friction. A poorly structured message slows everyone down. A candidate who cannot communicate clearly in an interview will raise questions about how they will communicate once hired. This does not mean you need to be perfect. It means you need to be intentional. Do the homework. Understand the company. Know the product. Prepare your examples. Structure your answers. Be clear on your motivations. Show ownership. Follow up properly. Make it easy for the team to trust you. Because when you interview for a remote or hybrid role, you are not only being assessed on your hard skills. You are being assessed on whether people can trust you when nobody is watching. And in the best teams, that bar is high… #RemoteWork #TechRecruitment #CandidateExperience #InterviewTips #CareerAdvice #Hiring #CommunicationSkills #Trust #FutureOfWork
-
Amine Kherbouche a réagi à ceciAmine Kherbouche a réagi à ceci16 TB of telco logs in. 367 GB on disk. A customer pulled that number up live on a call last week. One month of logs from their telco workload, sitting in object storage at roughly 2% of the raw size. And no, we don't get there by throwing data away. Zero sampling, full fidelity, every log line queryable. No dropped fields, no rollups. How it actually works: We store everything as Parquet, columnar layout, compressed with zstd. Logs compress absurdly well when you lay them out by column instead of by row. The same field repeated across millions of lines collapses to almost nothing. No fixed schema tax either. Elasticsearch makes you pay for indexing every field whether you search it or not. We index selectively, the fields you actually query, and leave the rest as cheap compressed columns you can still scan when you need them. Storage cost drops by orders of magnitude, and you keep 100% of the data. You don't trade your bill for your fidelity. When someone is staring at an Elasticsearch cluster that's 50x the size of the data that went into it, 16 TB landing at 367 GB is hard to argue with.
-
Amine Kherbouche a aimé ceciAmine Kherbouche a aimé ceciLFG! Having the worlds best company NVIDIA talk about Redpanda Data and NYSE makes me infinitely proud to partner with everyone at Redpanda. Live now.
-
Amine Kherbouche a aimé ceciAmine Kherbouche a aimé ceciLe dernier coup de gueule de Linus Torvalds pour la sortie de la version 7.1-rc4 du noyau Linux devrait faire réfléchir tous les RSSI et professionnels de la sécurité. Son constat ? La liste de sécurité de Linux est devenue "quasiment ingérable". En cause : une inondation de rapports générés par IA. Des centaines de personnes utilisent les mêmes outils automatiques pour remonter les mêmes bugs en masse, sans aucune vérification humaine ni compréhension du contexte. Comme le dit si bien Linus : « Ne soyez pas la personne qui envoie un rapport aléatoire sans réelle compréhension. » Quel est le rapport avec votre gestion des vulnérabilités ? Si votre stratégie de sécurité consiste à ouvrir les vannes et envoyer tous les rapports de scans bruts à vos équipes d'exploitation, vous créez la même crise : - Une surcharge cognitive pour vos ingénieurs, la fameuse "Alert fatigue", - Un "travail de fourmi" pour vérifier chaque alerte manuellement, - Des équipes paralysées par des tunnels de validation trop lourds. Détecter des failles est devenu facile. Savoir les requalifier, les centraliser et isoler les 5% critiques qui impactent réellement votre infrastructure, c’est là que réside la vraie gouvernance. La puissance de la détection n’a de valeur que si elle est guidée par du renseignement contextualisé et directement exploitable.
-
Amine Kherbouche a aimé ceciAmine Kherbouche a aimé ceciI wrote how we scaled the RIB (aka the routing table) in Akvorado (the flow collector we develop and use at Free/AS12322) to handle tens of millions of routes while keeping latency at minimum using "RIB sharding" (partition the RIB into several parts to enable parallel updates).
-
Amine Kherbouche a aimé ceciAmine Kherbouche a aimé ceciRecently, at the Betclic Group x Dev With AI meetup, I shared our journey on building and deploying AI agents specifically designed for Developer Experience (DX). We often talk about AI coding assistants, but the real power comes when you decouple the "agent" from the "chat interface." By deploying agents on our own infrastructure, we’ve enabled our users to consume AI through the tools they already use: 🔹 CLI: For instant, local execution within the terminal. 🔹 CI Pipelines: For automated logic verification and quality gates. 🔹 Webhooks: For event-driven workflows across our ecosystem. By treating AI agents as first-class citizens in our infrastructure, we move beyond simple productivity gains to building systemic guardrails. This allows us to scale development velocity while maintaining high engineering standards
-
Amine Kherbouche a aimé ceciAmine Kherbouche a aimé ceciThe team and I have been working hard. Striving for greatness. And it starts with delivering tangible results to our customers. Because of this GetVocal AI has been recognized in the VivaTech Top 100 Rising European Startups 2026. In the AI Voice and Conversational category, we built a solution around a Context Graph architecture that combines deterministic AI with generative models. It gives enterprises the control, transparency, and reliability they need to automate complex customer conversations. Many are only now starting to talk about these architectural foundations. While we built them three years ago. What makes me most proud isn’t the recognition, it’s the results our customers are seeing: real automation of complex CX conversations while keeping humans in control and improving outcomes. Again, I want to thank our valued customers for trusting us with such a critical part of their operations. And of course, our incredible team and investors (Creandum, Speedinvest, and Elaia) for believing in this vision from day one.
Expérience et formation
-
Teads
Voir toute l’expérience de Amine
Découvrez son poste, son ancienneté et plus encore.
ou
Recommandations reçues
1 personne a recommandé Amine
Inscrivez-vous pour y accéder