How an LLM Works: Tokens, Attention, Context, Evaluation, and Serving
How an LLM represents text → how a Transformer computes → how generation works → how we control context → how we evaluate → how we adapt the model → how we optimize and serve it.
Recent posts
Scan & Go Patent
Patent / Publication
Interview Preparation1
Interview / Algorithms
How Discord Stores TRILLIONS of Messages
Discord migrated a trillion‑plus message store from Cassandra (177 nodes) to ScyllaDB (72 nodes) in nine days with zero downtime. The success hinged on a Rust‑based data‑service layer for request coa…
Top 5 Most-Used Deployment Strategies
Deployment strategies trade off risk, downtime, and resource cost. The five most‑used approaches—Big Bang, Rolling, Blue‑Green, Canary, and Feature Toggles—each fit different release constraints and…
Top 7 Most-Used Distributed System Patterns
Distributed system patterns such as Ambassador, Circuit Breaker, CQRS, Event Sourcing, Leader Election, Pub/Sub, and Sharding provide reusable solutions for latency, resilience, scalability, and main…
10 Key Data Structures We Use Every Day
Data structures are the building blocks of everyday software, each offering distinct performance characteristics and cache‑friendliness. Choosing the right one—lists, arrays, stacks, queues, heaps, t…
Cache Systems Every Developer Should Know
Caching permeates every layer of a modern system—from CPU registers to browsers, CDNs, load balancers, message brokers, distributed stores, and databases. By keeping frequently accessed data close to…
Everything You Need to Know About DNS: Crash Course System Design #4
DNS is a hierarchical, decentralized directory system that translates human-readable domain names into machine-readable IP addresses through a three-tier chain of authoritative servers (root → TLD →…
But What Is Cloud Native Really All About?
Cloud Native is not simply “running your app on the cloud”; it is a blueprint for building web-scale applications that are more available, scalable, and agile, built on four pillars: microservices ar…
Debugging Like A Pro
Debugging is a disciplined, mindset‑driven process that starts with thorough information gathering and a reproducible environment, then proceeds through systematic investigation (print statements, de…
Top 5 Redis Use Cases
Redis’s in‑memory data structures make it ideal for high‑speed caching, session storage, distributed locks, rate limiting, and gaming leaderboards. Deploying Redis as a distributed cache or lock requ…
| CI/CD In 5 Minutes | Is It Worth The Hassle: Crash Course System Design #2 |
CI/CD automates building, testing, and deploying code so teams can ship higher‑quality software faster. CI is widely adopted and relatively straightforward, while true continuous deployment (CD) is p…
| Kubernetes Explained in 6 Minutes | k8s Architecture |
Kubernetes (k8s) is an open‑source container orchestration platform that automates deployment, scaling, and management of containerized workloads via a control plane (API server, etcd, scheduler, con…
CAP Theorem Simplified
CAP states that in the presence of a network partition a distributed system must sacrifice either Consistency or Availability. Real‑world systems often adopt hybrid or “best‑effort” strategie…