/ Blog
Writing
Notes on AI engineering, system design and full-stack development.
14 posts
- LLM Evaluation & Observability: The Skill Most AI Engineers SkipEveryone learns to call an LLM API. Almost nobody learns to measure whether it's actually working. Here's how to learn evaluation and observability before non-determinism bites you in production.6 min
- Agentic Workflows: How to Actually Learn to Build AI Agents'Agent' has become a marketing word. Here's the real learning path for agentic workflows — from a single tool-calling loop to multi-agent orchestration — with the failure modes nobody warns you about.7 min
- RAG for Developers: A Practical Learning Path From Zero to ProductionMost RAG tutorials stop at 'embed your docs and query them.' Here's the actual learning path — chunking, retrieval quality, evaluation — from someone who's shipped RAG pipelines in production.7 min
- Scalability vs Performance: The Distinction I Was MissingBreaking down why 'scaling' and 'performance' aren't the same conversation, and when to choose vertical vs horizontal scaling.2 min
- The System Design Fundamentals That Actually Get TestedInterviewers aren't scoring diagram complexity — they're testing a small set of fundamentals and your ability to reason through trade-offs.1 min
- Controlling a Running Claude Code Session From My PhoneAn active Claude Code session on your laptop can be monitored and steered live from your phone without the work ever leaving the machine.1 min
- Netflix's Architecture Story Isn't Really About MicroservicesThe most interesting part of Netflix's stack isn't the microservices at scale — it's Open Connect, their custom-built CDN.1 min
- Why the System Design Interview Bar Has Moved in 2026With AI writing boilerplate, architectural judgment and cost-awareness — not vocabulary — separate strong candidates now.1 min
- Certifications Aren't Dead in 2026, They're Just ConcentratedGeneric certifications have lost weight, but AI-specific credentials increasingly show up as preferred qualifications in real job postings.1 min
- Claude Code Is Becoming a Continuous-Use Tool, Not an Occasional OneAnthropic's temporary 50% higher weekly limits hint at Claude Code usage shifting from occasional tasks to continuous workflows.1 min
- Stop Building Portfolios Manually — Generate Them From Your GitHubYour GitHub already contains everything that matters — projects, tech stack, activity, proof of work. So just import your stuff in a structured JSON to build your portfolio, saving time and effort.4 min
- Integrating AI Into Web Apps Without the HypeMost AI integration tutorials skip the hard parts — latency, cost, fallbacks, and UX. Here's what I've learned shipping AI features in production.4 min
- My Go-To Next.js + TypeScript Stack for Full Stack AppsA breakdown of the exact stack I use when starting a new production-ready full stack project — from folder structure to database to deployment.3 min
- How I Cut API Latency by 40% — Lessons From ProductionThe specific techniques that actually moved the needle on API performance: query optimization, caching layers, connection pooling, and the mistakes I made along the way.4 min