About
Articles by Ehsan
-
Optimizing JavaScript Event Listeners for Performance
Optimizing JavaScript Event Listeners for Performance
While I am writing this, there are so many listeners are listening for an event on your website. Why we add them, it is…
Activity
10K followers
-
Ehsan Gazar shared thisThe most expensive line of code isn't the one that's slow. It's the one that makes two services impossible to deploy independently. Coupling is invisible until it isn't. Two teams shipping fine, then one change breaks the other's release. A hotfix that needs four sign-offs because four services share the same database schema. A "quick" refactor that turns into a 3-week coordination effort. This is the hidden cost of coupling — it doesn't show up in performance metrics. It shows up in your deployment calendar. **What coupling actually means:** Two things are coupled when a change in one forces a change in the other. That can be: - Shared database tables - Synchronous HTTP calls at runtime - Shared libraries with no versioning - Implicit assumptions about response shape **What autonomy costs:** More infrastructure. More duplication. Harder to enforce consistency. Teams that don't share code sometimes solve the same problem differently — and both solutions need maintaining. The trade-off isn't "coupled bad, autonomous good." It's: what's the cost of coordination at your current team size and release cadence? At 3 engineers, shared everything is fine. At 30, it's a bottleneck. At 300, it's existential. **The question to ask in every design review:** If this service needs to change, who else has to know? If the answer is more than one team — you've bought coupling. Make sure you meant to. → https://lnkd.in/eKJjy9hHHow to Defend Any System Design Decision in 6 QuestionsHow to Defend Any System Design Decision in 6 Questions
-
Ehsan Gazar shared thisMost engineers pick "consistency" without finishing the sentence. "We need strong consistency" gets said in design reviews constantly. Usually without anyone asking: at what cost? Strong consistency means every read sees the latest write — no matter which node responds. That sounds obviously correct. Until you think about what it costs. To guarantee it, your system coordinates across nodes before confirming a write. That takes time. Requires acknowledgment. And when a node goes down mid-write? You either block, or you roll back. Availability breaks. The CAP theorem formalises this: in a distributed system, during a network partition, you pick consistency or availability. Not both. What most engineers get wrong: they treat it as binary. In practice, most systems live in the middle. DynamoDB lets you choose consistency per-request. Cassandra tunes it per operation. Postgres with replication makes different promises depending on lag tolerance. The real question isn't "do we want consistency?" It's: which operations need it — and which can tolerate eventual? Payments? Strong consistency. User feed? Eventual. Inventory counts during a flash sale? Depends how angry your customers get when they buy something that's out of stock. This is the kind of trade-off analysis that separates senior engineers from staff engineers. Not knowing the answer — knowing which question to ask. https://lnkd.in/eKJjy9hHHow to Defend Any System Design Decision in 6 QuestionsHow to Defend Any System Design Decision in 6 Questions
-
Ehsan Gazar shared thisBig career moves rarely start with a five-year manifesto. They start when one bottleneck is named, visibility, scope, ownership, how your work shows up in reviews, and you stop improvising without a second pair of eyes. If you’re senior+, heavy on execution and light on influence, book a free 15-minute intro (conversation, not a pitch): https://cal.com/gazar What mentorship covers: https://lnkd.in/euQ9R8WN — Gaz
-
Ehsan Gazar shared thisEight services. Four teams. Release notes with a section called “deployment order.” They called it microservices. Sam Newman would call it something else: a distributed monolith. Same tight coupling, it just moved from function calls to HTTP. You keep the latency, the partial failures, and the observability tax without the one thing that justifies paying that bill: shipping one service without asking anyone else. Here’s the test I use: can this service hit production today, alone, no shared release train, no “wait for the other team”? If the answer is no, your next move isn’t another split. It’s finding the coupling and removing it. Patterns on the site (service boundaries): https://lnkd.in/ea2X2zjm The full issue, wrong boundaries (code size, org chart, technical layers), Newman’s deployability bar, Vernon’s “why split before how,” and what we learned splitting too early in fintech — is here: https://lnkd.in/ezD9UXZb — Gaz
-
Ehsan Gazar shared thisI built a commit message generator locally as a playground first—instead of reaching for a hosted API—so the model stays on my Mac (M4 these days), it costs nothing to run, and I never think about bills, rate limits, or someone else’s outage.Running Gemma 4 Locally for Your Next Feature: A Practical Decision Guide | GazarRunning Gemma 4 Locally for Your Next Feature: A Practical Decision Guide | Gazar
-
Ehsan Gazar shared thisThe Day 2 Problem hits different when AI writes your code. You get incredible velocity on Day 1—features ship at light speed. But three months later? You're staring at an 847-line component that nobody wants to touch. Not because it's broken, but because it was optimized for an imagined system, not yours. AI outputs locally perfect code that misses system-wide context: deprecated patterns, architectural constraints, why your team made certain decisions. The real shift: reviewing AI code is MORE demanding than reviewing human code, not less. You need fitness functions for your architecture. You need senior engineers reviewing critical paths. And you need to track how long it actually takes to onboard someone to that code. "It works" and "it's good" aren't the same thing. Maintainability shows up later. https://lnkd.in/eSjAmdjB #SoftwareEngineering #AI #CodeQuality #TechnicalDebt #DevCulture #EngineeringThe Day 2 Problem: Why Vibe Coding Is Creating a Maintenance Crisis and What Good Engineers Do Instead | GazarThe Day 2 Problem: Why Vibe Coding Is Creating a Maintenance Crisis and What Good Engineers Do Instead | Gazar
-
Ehsan Gazar posted thisAre AI tools making you a better engineer, or just a faster typist?
-
Ehsan Gazar shared thisYou’re not hitting AI limits — you’re wasting tokens. 💸 Every message you send to tools like Anthropic’s Claude forces it to reread your entire chat, which means what feels like progress is often just exponential cost and gradually worse output. 🔁📉 This isn’t a pricing problem. It’s a context hygiene problem. 🧠 The uncomfortable truth is that most engineers slow themselves down by sending too many messages, dumping too much context, and keeping chats alive far longer than they should. A better approach is simple but counterintuitive: start fresh chats more often 🧹, batch your thinking into a single well-structured prompt 📦, and only share the exact context that matters 🚫. The developers who understand this will move 10x faster ⚡ — while everyone else keeps blaming the pricing. 🤷♂️ #AI #SoftwareEngineering #LLM #Productivity #Startups17 Claude Code Token Management Hacks (Tier 1 to Tier 3) | Gazar17 Claude Code Token Management Hacks (Tier 1 to Tier 3) | Gazar
-
Ehsan Gazar reposted thisEhsan Gazar reposted this🚨 AI is making UX research faster, but not always better. I was reading this NN/g article by Maria Rosala on how research tools can hide serious methodological problems, and one point stood out: ❌ AI-integrated research tools can be especially misleading in usability testing. Why? Because usability testing depends on careful task design, neutral moderation, and correct interpretation of behavior. If an AI tool generates leading tasks, suggests where users should go, or frames questions in a way that primes participants (usability tools like Userology and TheySaid do that), the test stops reflecting real user behavior. ⚠️ That means the output may look polished, but the findings can still be flawed. A few risks really stood out to me: ❌ AI-generated tasks can unintentionally lead participants toward the “right” answer. ❌ Transcript-based analysis can miss the most important part of usability testing: behavior and user's silence ❌ Some tools blur the line between interviews and usability tests, which creates weak research and misleading insights. ‼️ The bigger issue is that automation can make poor research feel more credible. When a tool plans, moderates, and analyzes for you, it’s easy to trust the result without questioning the method behind it. AI can absolutely support research, but it should not replace research thinking. Speed is useful. Confidence is useful. But in usability testing, methodological rigor matters more than both. Worth reading: https://lnkd.in/e9S7SUXf
Experience & Education
-
Tipalti
View Ehsan’s full experience
See their title, tenure and more.
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Volunteer Experience
-
Secretary
Iranian Bushwalkers
- Present 7 years 4 months
Health
Publicly open to all nature-lovers, we are proudly the first bushwalking group established by Victorian residents of Iranian backgrounds. Affiliated by Bushwalking Victoria, Iranian Bushwalkers (IB) is a non-profit club whose main activities with insurance coverage includes Hiking, Bushwalking and Social Activities
-
Mentor | Community Manager
PersiaJS
Education
PersiaJS: A Dynamic Frontend Community for Engineers 🚀
🌐 Engage with a vibrant community of Engineers dedicated to Frontend excellence.
👩🎓 Students mastering Frontend, Backend, and DevOps skills through interactive learning.
💻 Code submissions reviewed by seasoned Mentors, fostering growth and improvement.
📚 Access downloadable resources, watch insightful tutorials, and seek guidance anytime.
🔍 Daily tips on cutting-edge techniques and tools shared on Linkedin and…PersiaJS: A Dynamic Frontend Community for Engineers 🚀
🌐 Engage with a vibrant community of Engineers dedicated to Frontend excellence.
👩🎓 Students mastering Frontend, Backend, and DevOps skills through interactive learning.
💻 Code submissions reviewed by seasoned Mentors, fostering growth and improvement.
📚 Access downloadable resources, watch insightful tutorials, and seek guidance anytime.
🔍 Daily tips on cutting-edge techniques and tools shared on Linkedin and Instagram.
📰 Explore a treasure trove of articles highlighting the latest industry insights. Join us on the journey of continuous learning and innovation! #PersiaJS #FrontendCommunity -
Mentor | Online Courses
Maktabkhooneh (مکتبخونه)
Education
Join thousands of students worldwide engaging in interactive and knowledge-packed courses, where learning transcends boundaries. Explore a diverse range of topics, from FrontEnd essentials to advanced technologies:
1️⃣ FrontEnd Mastery (HTML, CSS): Lay the foundation for stunning web development.
2️⃣ The Complete JavaScript Guide: Dive deep into the language powering the modern web.
3️⃣ React for Beginners: Unlock the potential of this powerful JavaScript…Join thousands of students worldwide engaging in interactive and knowledge-packed courses, where learning transcends boundaries. Explore a diverse range of topics, from FrontEnd essentials to advanced technologies:
1️⃣ FrontEnd Mastery (HTML, CSS): Lay the foundation for stunning web development.
2️⃣ The Complete JavaScript Guide: Dive deep into the language powering the modern web.
3️⃣ React for Beginners: Unlock the potential of this powerful JavaScript library.
4️⃣ React Advanced Techniques: Elevate your React skills to the next level.
5️⃣ AWS Concepts: Navigate the cloud with confidence and proficiency.
6️⃣ Clean Code Principles: Write code that stands the test of time and scrutiny.
7️⃣ NodeJS Essentials: Master server-side development with NodeJS.
8️⃣ React Native Fundamentals: Build cross-platform mobile applications seamlessly.
Experience an immersive learning environment, where you not only absorb knowledge but actively engage with the content. Ask questions, submit projects, and connect with a vibrant community of learners. Elevate your skills and embrace the transformative power of education. 🌐✨ -
Mentor
MAPSA HR & Training
- 9 months
Education
Nurturing Future FrontEnd Developers at Mapsa HR 🚀
I'm on a fulfilling journey, remotely guiding and mentoring students one-on-one in the art of FrontEnd development. It's an incredible experience, tailoring my teaching to each individual, empowering them to embark on their journey in the vibrant world of frontend technologies.
Join me on this exciting adventure, where personalized guidance meets the boundless potential of each student. Together, we're shaping the next generation…Nurturing Future FrontEnd Developers at Mapsa HR 🚀
I'm on a fulfilling journey, remotely guiding and mentoring students one-on-one in the art of FrontEnd development. It's an incredible experience, tailoring my teaching to each individual, empowering them to embark on their journey in the vibrant world of frontend technologies.
Join me on this exciting adventure, where personalized guidance meets the boundless potential of each student. Together, we're shaping the next generation of frontend developers and laying the foundation for successful careers. 🌐💻
Publications
-
Solving University Course Timetabling Problem (UCTP)
K. N. Toosi University of Technology
See publicationThis paper describes the University Course Timetabling Problem (UCTP). Timetabling describes a problem that usually arises in an educational context. At its core is the task of placing entities generically called events into a timetable made up of limited number of atomic units called time-slots. Different algorithms such as Linear, Genetic and Memetic are compared and described.
Explore more posts
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More