Software Architecture

A practical reference catalog of the eight architectures worth knowing — layered, modular monolith, hexagonal, event-driven, CQRS + event sourcing, microservices, serverless, and the strangler fig. Each with a diagram, the forces that make it the right call, the failure mode that makes it the wrong one, and a link to runnable reference code. Plus a decision flowchart so you pick on fit, not hype.

You can't ship a reliable LLM feature on vibes. Evals are the regression net for a dependency that's non-deterministic, drifts when the provider updates the model, and fails silently. How to build one without boiling the ocean: start with 30 real examples, layer three kinds of checks (assertion, LLM-as-judge, human), measure faithfulness, and run it on every prompt, model, and retrieval change.

Adding an LLM to your product is a distributed-systems problem with a non-deterministic dependency, not a single API call. When RAG actually helps (and when a prompt will do), how to think about vector databases and chunking without cargo-culting, the retrieval pipeline that separates demos from products, and the seven-point production checklist — evals, guardrails, cost ceilings, latency budgets, fallbacks, observability, and a human-in-the-loop boundary — to put in place before a real user touches it.

Naive RAG gets you a 70%-quality demo and a plateau. The gap to production is three retrieval levers most teams never pull: chunking on structure (not character counts), hybrid search (vector + keyword), and re-ranking an over-fetched candidate set. The deep dive on each, plus citations and the metrics that tell you where retrieval is failing. Retrieval quality beats model quality.

Events are a powerful tool and a terrible default. The three legitimate reasons to go event-driven, the anti-pattern that wrecks most implementations (events for request/response), the non-negotiables (idempotent consumers, dead-letter queues, versioned schemas), and choreography vs orchestration — with a decision rule so you reach for events when they earn their keep, not because a talk said microservices need a bus.

Three patterns separate systems that survive scale from systems that get paged at 3am. Cache invalidation and the stampede problem, idempotency keys done right, and retries with exponential backoff, jitter, and circuit breakers — plus how the three fit together into one reliability story. Get them right and most of your 3am pages quietly disappear.

Most architecture docs are write-once, read-never. ADRs are different because they capture the one thing that stays valuable: why a decision was made and which forces drove it. Five rules that make ADRs survive a busy team — keep them tiny, in the repo, record the forces and rejected options, write them at the right moment, and never edit a decided ADR (supersede it). Includes a minimal template.

The tenancy model is an early, hard-to-reverse decision that touches every query you write. Shared (one DB, tenant_id per row), siloed (a DB per tenant), and hybrid (shared for the long tail, siloed for enterprise) — the isolation/cost/complexity trade-off, the data-isolation and noisy-neighbor traps, per-tenant cost visibility, and why most startups should start shared but build the routing seam to earn their way to hybrid.

A comprehensive guide to software architecture for practicing engineers: what architecture actually is, core patterns (monolith, microservices, event-driven, DDD), making and recording good decisions, the most expensive mistakes, and how architecture changes as systems scale.

The same architectural mistakes repeat across hundreds of systems: premature microservices, tight coupling, ignoring operational concerns, and skipping documentation. These compound quickly—companies with architectural debt spend 60-80% of engineering time on maintenance. Learn the patterns, decision frameworks, and practical fixes.

Spent 5 days optimizing a function that runs once a day. Performance work without measurement is a hobby. Learn to measure first (profiling, APM), triage framework (frequency × impact × cost), high-leverage fixes (N+1, caching, indexes), and when optimization is overkill.

Missing auth check exposed every order in the system. Most incidents aren't Hollywood hacks—they're boring oversights. Learn the security mindset (hostile input, don't trust clients), 6 defensive practices, common web traps (XSS/CSRF/IDOR), and the hygiene checklist.

37 flags, no one knows what 18 do. Flag hell happens when power tools become toys. Learn the 4 flag types (release/experiment/ops/permission), lifetimes, gradual rollout patterns (5%→100%), cleanup discipline, and the avoid-flag-hell checklist.

Dashboard says everything's fine, but Slack is on fire. Observability isn't pretty charts—it's answering 'why is THIS user broken?' Learn logs/metrics/traces workflow, correlation IDs, structured logging, the 15-min debugging playbook, and anti-patterns.

That 40-page doc no one opens wastes weeks. Good docs aren't comprehensive—they answer the right question in 10 minutes. Learn the 4 doc types (READMEs, architecture, runbooks, ADRs), templates, discoverability tactics, and the upgrade checklist.

That user_id column you added in 5 minutes now blocks sharding, mergers, and compliance. Database mistakes hide quietly then surface when you're growing fast. Learn the 6 mistakes (God Tables, wrong IDs, missing indexes, mixed workloads) and the pre-launch checklist.

That 'quick' API you shipped 5 years ago now blocks every change. APIs are promises—bad ones compound pain. Learn tool-agnostic principles, REST/gRPC/GraphQL/tRPC trade-offs, TypeScript patterns, versioning strategies, and the pre-merge checklist for APIs that last 10 years. Updated for 2026.

Your AWS bill just hit ₹15 lakhs/month and is eating your runway. Learn how to cut cloud costs by 40-60% in 30 days without killing velocity—from quick wins to building a FinOps culture in small teams.

Your dashboard query just crashed production. Learn when to separate OLTP from OLAP, how to introduce a data warehouse, and whether you actually need event streams and a lakehouse—with a clear maturity checklist.

Stop building platforms too early or waiting too long. Learn the exact signals that indicate you're ready for platform engineering, what to build first, and how to show ROI without creating an ivory tower.

Stop chasing coverage numbers. Learn how to design right-sized testing strategies that evolve from MVP to enterprise—protecting critical flows without slowing your team down. Updated for 2026 with TypeScript, Vitest, Playwright, and AI-assisted testing.

Big rewrites fail. Learn how to modernize legacy systems incrementally using the strangler fig pattern: map dependencies, prioritize modules, use feature flags for gradual rollout, and migrate data safely—all while shipping features.

Move from emotional arguments about 'we must refactor everything' vs 'just ship' to a clear decision framework. Learn when to pay down debt, when to accept it, and how to speak the language of business trade-offs.

Learn to identify and eliminate code smells that silently degrade software quality. Practical guide with Python examples showing how to recognize warning signs and implement systematic prevention strategies.

Transform cryptic code into clear, maintainable systems. Learn practical clean code principles through real Python examples that enable teams to scale from startup to enterprise.

Transform code reviews from reactive bug-hunting into proactive system building. Master strategic review practices that scale engineering excellence and build maintainable systems.

The DRY principle seems simple on the surface: don't write the same code twice. But when should you eliminate duplication? When is duplication actually beneficial? Learn the complete guide with practical examples.

SOLID principles aren't just academic concepts—they're practical guidelines that solve real problems every developer faces. Learn each principle through relatable examples and see how to apply them effectively.

A practical guide to choosing between microservices and monolithic architectures. Real-world case studies and decision frameworks for Indian companies.

A comprehensive framework for technology stack decisions. Learn how to evaluate options, avoid common pitfalls, and choose technologies that scale with your business. Updated for 2026 with TypeScript-first approaches, AI tooling considerations, and modern runtime options.

The architectural principles that actually work in practice: design for change not perfection, treat data ownership as an architecture decision, use Conway's Law deliberately, build observability in from day one, and evolve incrementally using the Strangler Fig pattern. From leading architecture at companies scaling from 10 to 150 engineers.

Beyond the buzzwords: practical clean code principles that improve maintainability and team productivity.

Battle-tested architecture patterns for Indian startups scaling from thousands to millions of users. Practical implementation strategies with real-world examples.