Göteborg, Västra Götalands län, Sverige
230 tn följare
Fler än 500 kontakter
230 tn följare
Fler än 500 kontakter
Info
I am a senior software developer and architect with over 20 years of experience in…
Aktivitet
230 tn följare
-
Kasra Jadid Haghighi har delat dettaSilence Is Not Neutral When People Are Being Killed in Iran I haven’t been able to stop thinking about Iran these past days. Not as news. Not as politics. As people. According to reports from Iranian activists and opposition networks, thousands of protesters were killed within just a few days during the most recent protests. This happened during an extreme crackdown by the Islamic Republic regime, while severe internet restrictions made independent verification almost impossible. When civilians are killed in such numbers and in such a short time, there is no other honest word for it. It was a massacre. At the same time, tens of thousands have reportedly been arrested. Many detainees face torture, forced confessions, and unfair trials. Executions and death sentences are used to spread fear. Internet shutdowns are imposed to hide the scale of what is happening. These are not random events. They reflect a system that is willing to use extreme violence against its own people to stay in power. I am not an expert, an activist, or a politician. I am just someone who believes that silence is not neutral. When we look away, we make it easier for this kind of violence to continue. I don’t live in Iran and I can’t be on the streets. But I can help share what Iranians themselves are saying. I can keep talking about Iran even when the world moves on. I choose not to be silent. #FreeIran
-
Kasra Jadid Haghighi har delat detta.NET Tip: Keyed Services in .NET 8+ One of the useful additions in modern .NET is Keyed Services. It allows you to register multiple implementations of the same interface and resolve them by a specific key. This becomes really practical when you want to separate behaviors but still keep a clean architecture. For example, you can use it for: • Factory patterns where you select the right implementation based on a key • Multi-tenant setups where each tenant has its own configuration or logic • Load balancing or routing between different strategies • Named or static modes such as “default”, “fast”, or “safe” The nice thing is that it’s all supported natively by the built-in dependency injection container, so there’s no need for extra libraries or complex factory logic. If you often deal with scenarios where one interface has several specialized implementations, this feature can really simplify your codebase. Code on GitHub: https://lnkd.in/drMezVCc
-
Kasra Jadid Haghighi har delat detta.NET Tip: One Place for All Your NuGet Versions Keeping NuGet package versions aligned across multiple projects can be challenging, especially as a solution grows. Central Package Management (CPM), introduced in .NET 6 SDK, helps solve this by defining all package versions in a single place called Directory.Packages.props. This file sits at the root of the solution and acts as the central source of truth for package versions. Each project then references the packages by name only, without specifying a version. The result is consistent dependencies, easier updates, and fewer version conflicts across all projects in the solution. Code on GitHub: https://lnkd.in/diNbY6jg
-
Kasra Jadid Haghighi har delat dettaThis is a hands-on approach to teach z-index to CSS learners 😁 #HappyFriday
-
Kasra Jadid Haghighi har delat dettaBefore Writing Code, Discover the Domain Many software projects start by jumping straight into code. We create controllers, entities, or even microservices before we really understand the domain. But if we skip discovering the domain boundaries first, we make things much harder later. Let’s take a simple example: a bookstore. At first, a “Book” class sounds easy to model. We think of fields like Title, Author, Price, and Pages. But once we look closer, we see that “Book” means different things in different contexts. In the Store context, a book has a price, a discount, and a category. In the Warehouse context, the same book has a quantity, a shelf number, and a supplier. These are different domains with their own rules and responsibilities. If we put everything into one big Book class, we end up mixing things that don’t belong together. The result is a complex and confusing model that is hard to maintain and understand. It also becomes difficult for multiple teams to work on the same codebase. We do not need to start by creating microservices immediately. Instead, we can separate bounded contexts inside one solution. For example, we can have different projects such as Store.Domain and Warehouse.Domain. Each can have its own application services and logic. We can even use a single database, but with different schemas for each context. One of the most practical ways to discover these domains and boundaries is through event storming. It helps teams visualize the flow of events in the system and identify where one domain ends and another begins. It is a great collaborative tool for building a shared understanding of the system. This approach gives us a clean and modular system that is easy to maintain and extend. It also fits well with the Agile way of working, where the project grows step by step. Later, if needed, each context can become its own independent microservice and use a separate database. When we start a software project, discovering the domain boundaries is one of the smartest things we can do. It saves time, reduces complexity, and helps teams work more efficiently.
-
Kasra Jadid Haghighi har delat dettaC# Tip: Pattern Matching for Clarity and Safety Pattern matching helps you describe what you expect in a simple and direct way. It reduces the need for extra null checks, type casts, and deeply nested conditions. The result is code that’s easier to read, maintain, and reason about. It’s a small feature, but one that makes everyday logic a bit cleaner and more consistent.
-
Kasra Jadid Haghighi har delat dettaC# Tip: Async/Await couldn’t be explained better! 😂 This kid on a bike is actually your await SomeAsyncMethod() when ConfigureAwait(true) is on. He jumps off the bike (awaits), drops off the package (does some async work), then gets back on the same bike to keep going (resumes on the original context). Now here's a question for you 👇 What do you think happens if you use ConfigureAwait(false) instead? I’ll leave the answer in the comments! #HappyFriday
-
Kasra Jadid Haghighi har delat dettaBuilding reliable AI agents: 8 questions developers are asking After my last post about Parlant, I got a lot of interesting questions. I thought it could be useful to share some of the most common ones and my answers here, since many of us as developers face the same challenges when trying to build trustable AI agents. For those who didn’t see it, Parlant is an open-source framework that tackles one of the hardest problems with AI agents: reliability in sensitive workflows. Most of the time, agents built only on top of large prompts or rigid flows either hallucinate, lose track, or break when the conversation gets more complex. Parlant’s approach is to give developers proper building blocks like guidelines, journeys, and canned responses so you can actually design and test agent behavior instead of just hoping the model behaves. Here are some of the questions I’ve been asked since my last post: Q1: Is Parlant just another chatbot framework? A1: Not really. Chatbot frameworks are usually rigid flows, and pure LLM bots are unpredictable. Parlant sits in between: you define rules (Guidelines) and flows (Journeys), and the LLM works inside that structure. Q2: Do I need a specific model to use it? A2: No. It’s model-agnostic. I tested it with LLaMA locally, but you can connect it to OpenAI, Anthropic, or others. Q3: How hard is it to get started? A3: Pretty simple. A few lines of Python and you can define your first agent with guidelines. The Quickstart doc makes it easy. Q4: What’s the difference between Guidelines and Journeys? A4: Guidelines are rules like “always verify identity first.” Journeys are like state diagrams that define the flow of a conversation, but they can adapt if the user goes off track. Q5: What if I need the agent to use exact phrasing? A5: That’s what Canned Responses are for. You can predefine what the agent should say in sensitive moments so there’s no risk of hallucination or wrong wording. Q6: Can I connect Parlant agents to APIs or tools? A6: Yes. You can connect it to APIs or define custom tools in Python, and Parlant controls when they’re called. The LLM doesn’t get to trigger them on its own. Q7: How do I debug if something goes wrong? A7: Parlant has a Playground where you can see which guideline fired, what state the agent is in, and which tool was triggered. This makes it much easier to test compared to black-box prompting. Q8: Is it production-ready? A8: From what I’ve seen, yes. It’s open-source and already being used in areas where compliance matters, like finance and healthcare. Of course, you still need to test it in your own environment. I will share the link of my previous post and Parlant's GitHub in the comments.
-
Kasra Jadid Haghighi har delat dettaWhy We Over-Engineer (and How to Avoid It) I’ve often seen how easy it is to slip into over-engineering. Sometimes it’s because we want to “future-proof” everything. We imagine features or scenarios that might happen one day, and we start adding complexity just in case. But in reality, this often breaks with the spirit of agile: deliver what is needed today, in a way that is still flexible for tomorrow. That’s where design patterns and SOLID principles can help, writing code that is both maintainable and adaptable, without overcomplicating things. Other times, over-engineering happens for a completely different reason: we just want to learn something new. I’ve seen (and been tempted myself) to implement things like CQRS or even Event Sourcing in a simple CRUD project, not because the project needed it, but because I was curious. While curiosity is great, the downside is that it can leave the team with unnecessary complexity, fragile implementations, and long-term maintenance costs. I’ve learned that if you want to practice a new concept, it’s better to create a side project, a playground repo, or even small experiments. That way, you get the learning without burdening your team or product with complexity that doesn’t serve real requirements.
Kasra Jadid Haghighi har kommenterat ett inlägg
3 mån
Congratulations! 🎉
Kasra Jadid Haghighi har svarat på en kommentar
7 mån
Deepak K N V D L It should be generated by AI I think. Maybe they used "a baby" in the prompt instead of "a child" :))
-
Kasra Jadid Haghighi har gillat dettaKasra Jadid Haghighi har gillat dettaFor the last two years, investors have treated AI like a technology cycle. New models. New chips. New startups. New enterprise software. New valuations. But that framing is now too small. AI is no longer just lifting a few software companies or creating excitement around Nvidia. It is becoming one of the main forces carrying the broader economy. And that makes the opportunity bigger. But also the risk. Because if this AI investment cycle breaks, it will not look like the dot-com crash.... //not an investment advice, only my personal opinion. Please do you own research.The AI Boom Is Becoming the Market Story headed For A Reckoning, are you ready for it?The AI Boom Is Becoming the Market Story headed For A Reckoning, are you ready for it?
-
Kasra Jadid Haghighi har gillat dettaKasra Jadid Haghighi har gillat dettaVi söker en Engineering Manager till Winningtemp i Göteborg 💜 Den här personen kommer att jobba tätt ihop med mig i vårt tech leadership team - så för mig handlar rekryteringen lika mycket om ledarskap och teknisk höjd som om att hitta en kollega jag kan lära mig av och bolla det svåra med. Någon som vågar utmana, som lyssnar, och som vill bygga på riktigt. Kanske viktigast av allt: att vi har kul tillsammans! Rollen handlar om att utveckla människor, äga vår utvecklingsprocess och driva AI-transformationen i vår tech-org - bland annat hur vi använder Claude Code och agentic workflows i vardagen. Rollen är baserad i Göteborg, hybrid. Hör av dig eller tipsa någon du tror på. Länk i kommentarerna 👇
-
Kasra Jadid Haghighi har gillat dettaKasra Jadid Haghighi har gillat dettaMost people still think the biggest challenge in enterprise AI is the model. I don’t think so. The harder question is: Can your infrastructure stay secure, fast, and continuously available when AI starts touching mission-critical systems? That matters because enterprise AI is shifting from copilots and demos to agentic systems that interact with live business data, support real-time workflows, and in some cases take actions autonomously. A few key takeaways: 1. Enterprise AI is becoming operational, not experimental This is no longer just about generating correct answers or photo-realistic images. It is about supporting systems in finance, logistics, ERP, telecom, customer operations, and other environments where downtime is expensive. 2. Reliability is now part of the AI stack As AI gets embedded deeper into operations, uptime, failover, and recovery matter just as much as model quality. A few minutes of disruption may have been acceptable before. When AI systems are making decisions, seconds matter. The bar for downtime is rising quickly. 3. Security has to move closer to the data Once AI agents start working across sensitive systems, the risks grow: • over-permissioned agents • data leakage • weak visibility controls • inconsistent governance across apps and workflows 4. Post-quantum readiness is becoming real More companies are starting to plan for a world where attackers can capture encrypted data today and decrypt it later with more advanced quantum systems. 5. The winners may not be the ones with the flashiest model They may be the ones with the most resilient infrastructure: • no downtime • faster recovery • tighter security • better control over live enterprise data Oracle’s latest announcement is interesting Read more about it here: https://lnkd.in/gFzAUtQm What caught my attention is not just that Oracle launched another database update. It is what the update signals. Maximum Availability Architecture that delivers: Platinum-tier availability with failover typically under 30 seconds Diamond-tier availability for ultra-critical workloads with failover typically under 3 seconds Deep Data Security at the data-layer for AI-driven access support for post-quantum cryptography That is a useful signal of where enterprise infrastructure is heading in the AI era. The next AI battleground is not just intelligence. It is trust, uptime, security, and control. Oracle #oraclepartner #ad #ai #technology #innovation
-
Kasra Jadid Haghighi har gillat dettaKasra Jadid Haghighi har gillat dettaBuild an AI Employee with Claude! How to become a Full Stack AI Engineer in 2026.Build an AI Employee with Claude! How to become a Full Stack AI Engineer in 2026.
-
Kasra Jadid Haghighi har gillat dettaKasra Jadid Haghighi har gillat dettaEveryone wants AI now. And it is costing them millions. But, What breaks if you kill the pilot? For many companies, the answer is: not much. That is not just my opinion. The data is starting to say the same thing. RAND found that more than 80% of AI projects fail, roughly twice the failure rate of non-AI IT projects. MIT NANDA found that despite $30–40B in enterprise GenAI investment, only 5% of integrated AI pilots are extracting major value, while most show no measurable P&L impact. S&P Global found that the share of companies abandoning most AI initiatives rose from 17% to 42%, with the average organization scrapping 46% of AI proof-of-concepts before production. BCG found that only 5% of companies are achieving AI value at scale, while 60% are seeing minimal revenue or cost gains despite serious investment. Different reports. Same underlying truth. AI does not magically fix how a business actually works. If anything, it exposes the cracks: ↳ Data lives in twelve places. ↳ Metrics mean one thing in finance and another in sales. ↳ Ownership is passed around like a ping-pong ball. ↳ Governance shows up after the demo, not before it. ↳ Responsible AI is treated like a slide, not an operating system. Still, many companies keep launching pilots and hoping AI will smooth over the mess. But tools do not create transformation. Operating models do. The teams actually getting value do three things differently: 1. They use AI to expose the broken parts of the business. Not before transformation. Through transformation. The failed prompts, messy handoffs, bad data, duplicate processes, and unclear approvals are not distractions. They are the map. 2. They kill pilots early. If a pilot cannot connect to revenue, cost, risk, speed, customer experience, or decision quality, it should not survive just because it sounds innovative. Fail fast is not failure. Keeping a useless pilot alive is. 3. They use AI for the right job. If the goal is simple automation, there are often cheaper and easier ways to do it. AI is powerful when the work requires language, reasoning, context, decision support, creativity, or judgment at scale. Not every workflow needs an agent. Not every problem needs a model. The test is simple: Turn it off. If no one notices, it was never transformation. It was theatre.
-
Kasra Jadid Haghighi har gillat dettaKasra Jadid Haghighi har gillat dettaEnterprise AI is moving fast, but most companies are discovering the hard part is no longer just choosing the right model. It is making AI reliable, scalable, cost-efficient, and production-ready. In this LinkedIn Live, I’ll be joined by Ben Boren, CTO & Co-Founder of SwarmOne, for a practical conversation on what it really takes to move enterprise AI from experimentation to real infrastructure. We’ll discuss: → Why AI infrastructure has become one of the biggest bottlenecks in enterprise adoption → The rising cost of inference and why GPU utilization matters more than ever → How enterprises should think about multi-model, multi-tenant, and agentic workloads Why heterogeneous silicon, disaggregated inference, and smarter orchestration may define the next phase of AI infrastructure What real production teams need beyond demos, benchmarks, and hype This will be a practitioner-to-practitioner conversation, focused on the real challenges companies are facing as they try to scale AI beyond pilots. Join us live for a clear look at where enterprise AI infrastructure is heading, and what teams need to get right before AI can truly move into production.
-
Kasra Jadid Haghighi har gillat dettaKasra Jadid Haghighi har gillat detta#IBMPartner The conversation that will define the next five years of business is happening in Boston this May. It’s happening right now. Think 2026 is taking place May 4–7, and honestly, this is one of those moments where what gets discussed shapes everything that comes next. As someone who's spent a lot of time at the intersection of data and decision-making, I can tell you: the move toward agentic AI is a fundamental redesign of how organizations operate. And Think 2026 is where that conversation is happening at the highest level. You've got 130+ thought leaders, 7 keynotes and people flying in from 80 plus countries. But what's really caught my attention are the announcements coming out of this event. ⭐ IBM Bob (ibmcreator.com/4eZUhY6) is genuinely exciting. It’s an agentic development partner that works across the entire software lifecycle. The 45% productivity gains and near-90% faster delivery at client organizations aren't numbers you can ignore. ⭐Quantum (ibmcreator.com/49dA90V) represents the next frontier. Organizations are already seeing practical applications (from drug discovery to optimization problems that classical systems can't solve efficiently). This is where competitive advantage gets built. ⭐ IBM Concert (ibmcreator.com/4usBpW8) tackles smth I see holding organizations back constantly – the inability to act on insights in a coordinated way. It correlates signals across infrastructure, apps and networks into a single view, then helps teams move from insight to action. ⭐ On the data side, Confluent integration (ibmcreator.com/4dlOQ4y) addresses a critical gap: mainframe data that's locked away and delivered in batches. Now it can flow continuously into real-time AI and analytics environments. That's transformative for fraud detection and financial services. ⭐ Sovereign Core (ibmcreator.com/48DL4Rm) is strategically important for regulated industries. It embeds governance at the infrastructure level. Compliance becomes foundational. Now, I know not everyone can make it to Boston, and that's completely fine. IBM is livestreaming the keynotes and spotlight sessions, so you can tune in from wherever you are: https://lnkd.in/gxjTMZrr These conversations don't happen every day, and trust me, you'll want to hear what's being said. If you could ask one question to the leaders speaking at #Think2026, what would it be?
-
Kasra Jadid Haghighi har reagerat på dettaKasra Jadid Haghighi har reagerat på dettaCustomer support hasn't changed in 30+ years because nobody solved the math. AI just did. You call your telecom provider because the internet's down. You wait 15 mins. An agent picks up, puts you on hold again, comes back with “Have you tried unplugging it?..” You're now 27 mins in and even angrier than when you started. This is the reality for millions and millions of customers every day. And here's the brutal part: it's not about a lack of care. A company like Deutsche Telekom or Deliveroo handling millions of customers simply can't staff enough humans to answer every billing question, process every return, handle every outage notification in real time. So the system breaks, and everyone loses. Turns out, the fix was a voice. Or voice agents, to be precise. These agents handle order tracking, billing disputes, password resets, identity verification for account recovery (the stuff that makes up 60 percent of call volume). They respond in sub-second time. They work 24/7. They speak 70-plus languages. And when they hit smth complex… they hand off to a human with full context already loaded into your CRM. ElevenLabs' ElevenAgents platform is built specifically for this. You upload your SOPs, your knowledge base, your policies. The agent learns from that. E-commerce companies use it to generate return labels and suggest alternatives based on purchase history. Financial services use it for identity verification and dispute escalation with full audit trails, etc. Right now, they're processing over 250 million (!!) conversations every month. Revolut, Meesho, Immobiliare, Cisco – these aren't startups. These are companies handling MASSIVE volume. If you want to see what this actually looks like, their agents platform is worth a proper look: https://lnkd.in/gK-h2Xks Uncomfortable question: would you sacrifice your own time so a support agent could keep their job?
Erfarenhet och utbildning
-
Winningtemp
Se Kasras fullständiga erfarenhet
Se titel, anställningstid med mera.
eller
Licenser och certifikat
Språk
-
Persian
Modersmåls- eller tvåspråkig nivå
-
English
Professionell yrkeskunskap
Mottagna rekommendationer
2 personer har rekommenderat Kasra
Gå med nu för att se