PinnedInJavarevisitedbyRajesh Kumar·Sep 4Your Query Took 12 ms. It Can Still Hurt Production.The four EXPLAIN (ANALYZE, BUFFERS) signals I check before trusting PostgreSQL query latency
PinnedInLevel Up CodingbyRajesh Kumar·Aug 22CQRS: Architecture Pattern You Probably Don’t Need — Until You Really DoWhy separating reads from writes can unlock scale — and why doing it too early can make your system worse.
PinnedInLevel Up CodingbyRajesh Kumar·Aug 298 Coding Patterns That Turn Good Code Into Good ArchitectureThe real architecture of a system lives in its state models, boundaries, failures, retries, naming, and everyday code changes.A response icon2A response icon2
Rajesh Kumar·Oct 5RAG and Enterprise Search: What Makes an Answer Trustworthy?Finding relevant information is only the first step.
Rajesh Kumar·Oct 4Beyond the Prompt: Six Building Blocks of Production LLM SystemsA refund assistant shows where reliability, latency, and cost become architectural decisions.
Rajesh Kumar·Oct 4Structured Outputs: Your AI Returned Valid JSON. Can You Trust It?Predictable responses are the start. Your application still owns correctness.
Rajesh Kumar·Oct 4Prompt Hierarchy and Who Gets to Instruct the AIHow AI resolves conflicting instructions — and why documents must not take control of your assistant.
Rajesh Kumar·Oct 3Pre-Training vs Fine-Tuning: What Are You Actually Teaching the Model?Choose the training method by identifying what the model is missing.
In0xAIArchitectbyRajesh Kumar·Sep 29Designing a Rate Limiter: The Algorithm Is the Easy PartA practical system-design journey from counters and token buckets to shared state, atomicity, and failure handling.
InStackademicbyRajesh Kumar·Sep 24I Chose Databases for a College Guest Lecture. I Started With Failure Modes.A practical path from row locks to caching, Redis, NoSQL, and vector search — starting with application behavior, not database products.