Welcome to Algomi Blogs
Who this is for, how to use it to prepare for a Solutions Architect interview, and how a post gets published here.
Created 2026-09-15 · Updated 2026-10-12 · System Design
Algomi Blogs
System design, AI engineering, databases, networking, concurrency and more, written for engineers building real trading and fintech systems — part of the Algomi Suite.
Browse by topic (29)
AI Engineering (7)Java (7)Distributed Systems (6)Reliability (6)Security (6)AI Agents (5)Caching (5)Concurrency (4)Databases (4)Low-Level Design (4)Messaging (4)Networking (4)Streaming (4)APIs (3)Architecture (3)Consistency (3)Design Patterns (3)Harness Engineering (3)Observability (3)RAG (3)Transactions (3)Agent Memory (2)Capacity Planning (2)Enterprise Architecture (2)Fintech (2)Kafka (2)Machine Coding (2)Storage (2)Tool Use (2)
The Enterprise Architect's Interview Framework: From Business Problem to a Design You Can Defend
Start here if you are preparing for an Enterprise Solutions Architect interview: the ten-step framework, the landscape you design into, the four diagrams to draw, how to speak about trade-offs, and a worked example.
Created 2026-10-12 · Updated 2026-10-15 · System Design
System Design Made Simple
An original system design series — five foundation guides, fifteen focused case studies, and two deep dives — start here.
Created 2026-09-15 · Updated 2026-10-12 · System Design
System Design Starts with a Library, Not a Server
Follow one school library's reservation app from a single server to forty schools: requirements, estimation, scaling, storage choices, replication, consistency, CAP/PACELC, and the hashing tricks behind large-scale systems. Every idea arrives because a real problem forces it. Guide 1 of 5.
Created 2026-09-15 · Updated 2026-10-15 · System Design
What Happens Between a Click and an Answer?
Follow Riya's tap on 'Reserve' from her phone to the library's database and back: IP, DNS, HTTP, TCP/UDP, CDNs, load balancers, proxies, API styles and live updates, each introduced by the problem that forces it. Guide 2 of 5.
Created 2026-09-15 · Updated 2026-10-15 · System Design
Let Each Part Do Its Job: How One School Library App Grew Up
Follow one school library's reservation app from a single program to a system with a broker, outbox, saga and read model. Every pattern arrives because a real complaint forces it: monoliths vs microservices, async jobs, queues and streams, delivery guarantees, retries, outbox, sagas, CQRS, event sourcing, and the object patterns inside the code. Guide 3 of 5.
Created 2026-09-15 · Updated 2026-10-15 · System Design
A Good System Still Works on a Bad Day
Meera's library app has a very bad term: a dead server, a lost zone, a runaway retry storm, a deleted table and a stolen reservation. Each disaster introduces one idea: SLIs/SLOs, redundancy and failover, circuit breakers and rate limiting, authentication vs authorization, OAuth/OIDC, and recovery targets. Guide 4 of 5.
Created 2026-09-15 · Updated 2026-10-15 · System Design
Explain the Design, Then Defend the Decisions
Arjun, who built the library app in guides 1 to 4, sits a system design interview and is asked to design... a library reservation system. Follow his 45 minutes: scope, data, API, the last-copy race, lost responses, failures, growth, then two follow-up cases (async reports, a live feed), and the two-week plan he used to prepare. Guide 5 of 5.
Created 2026-09-15 · Updated 2026-10-15 · System Design
Hotel Reservations: How to Promise a Room Only Once
From one guesthouse in Goa to 80,000 hotels: how a booking system claims every night of a stay exactly once, keeps payment uncertainty away from inventory, and stops two guests winning the same room. Includes the three follow-up questions interviewers use to separate good answers from great ones.
Created 2026-09-16 · Updated 2026-10-15 · System Design
Airline Management: One Journey, Several Carefully Kept Promises
Follow Ananya and her father from Pune to Srinagar via Delhi, through a fog cancellation and an aircraft swap, to see how an airline system models dated flights, multi-leg bookings and disruptions as durable workflows instead of simple status edits.
Created 2026-09-16 · Updated 2026-10-15 · System Design
Inventory Management: Knowing What You Can Still Promise
Back-to-school season at an online stationery shop: why on-hand stock isn't available stock, how atomic reservations stop two orders claiming the same calculator, and how shipments, cancellations and warehouse transfers stay honest.
Created 2026-09-16 · Updated 2026-10-15 · System Design
ATM Design: When Software Must Agree with a Physical Machine
A monsoon power flicker hits a village ATM just as ₹10,000 slides out. How do you design a withdrawal that survives a network failure between the bank's authorization and the machine dispensing cash, when cash can never be rolled back?
Created 2026-09-16 · Updated 2026-10-15 · System Design
Designing Authentication: Proving Who You Are
A stolen laptop, a forgotten password and a 'Sign in with Google' button: follow one student's account to design registration, login, sessions, recovery, MFA and OIDC sign-in around durable identity instead of trusted client claims.
Created 2026-09-16 · Updated 2026-10-15 · System Design
Designing a CDN: Put Popular Content Near Its Readers
A national e-textbook platform, results day, and one leaked answer key: how a CDN serves popular content close to readers, why the cache key is a correctness decision, and how to stop a stampede from flattening the origin.
Created 2026-09-16 · Updated 2026-10-15 · System Design
Designing Photo Storage: From Small Files to a Large Photo Service
Sunita uploads 400 wedding photos from a train with patchy Wi-Fi. Follow them through a photo service that separates bytes from metadata, treats upload, processing and deletion as retryable workflows, and borrows one big idea from Facebook's Haystack.
Created 2026-09-16 · Updated 2026-10-15 · System Design
Designing a Job Scheduler: Remember the Work, Expect Retries
Every Monday at 08:00, 600 schools send progress reports to parents. One Monday some parents get two and some get none. Here is why a job definition, a scheduled run and an execution attempt must be three different records, and why leases don't erase duplicate effects.
Created 2026-09-16 · Updated 2026-10-15 · System Design
Designing a Professional Network: People, Connections, and Feeds
Follow Imran, a junior developer looking for his next job, through a LinkedIn-like network: how to model connections and follows, why search and feeds are projections, and how a hybrid fan-out keeps one famous post from flooding the system.
Created 2026-09-16 · Updated 2026-10-15 · System Design
Designing an Online Chess Game: Rules First, Servers Second
An inter-school blitz final between Chennai and Delhi tests everything: why the server must be the referee, why a board picture isn't the full state, and how one game stays correct through double taps, a ticking clock and a server failover.
Created 2026-09-16 · Updated 2026-10-15 · System Design
Designing BookMyShow: Selling Each Seat Only Once While Thousands Rush In
9:00 on a Friday, first-day-first-show bookings open for a superstar's film, and 100,000 fans want 400 seats. How a ticketing system holds seats for a few minutes, keeps payment uncertainty away from inventory, and stays fair when the crowd arrives.
Created 2026-09-26 · Updated 2026-10-15 · System Design
Designing YouTube: From One Upload to a Video Anyone Can Play
A 17-year-old uploads a 2 GB cooking video over a shaky connection. By evening it has a million views, including his grandmother's, on a 2G phone. Follow it to see how a video platform accepts huge uploads, builds many playable versions, serves them from nearby caches, and counts views without a slow database.
Created 2026-09-26 · Updated 2026-10-15 · System Design
Designing Netflix: Putting the Movie in a Box Near the Viewer
Friday, 9 pm: a hit series drops its new season and a family in Kochi presses play along with millions of others. How a streaming service keeps the cloud for decisions and puts the video bytes on cache appliances inside viewers' internet providers, stocked overnight before anyone asks.
Created 2026-09-26 · Updated 2026-10-15 · System Design
Designing Hotstar: Live Cricket for Tens of Millions at Once
World Cup final night: a student on a cheap phone in Bihar and a family on a big TV in Pune press play at the toss, along with tens of millions of others. How a live-sports platform survives a crowd that arrives all at once: a stream for every screen, caching what can be cached, scaling in steps before the toss, and switching off extras so the video keeps playing.
Created 2026-09-26 · Updated 2026-10-15 · System Design
Designing a Payment System: Charging a Customer Once, Even When Networks Fail
Mrs. Rao pays ₹1,000 for Diwali groceries, the network hiccups, and she taps Pay again. Follow her payment through idempotency keys, an honest 'unknown' state, webhooks, a double-entry ledger and daily reconciliation, then up to the enterprise questions: stakeholders, compliance, failover and how the design grows.
Created 2026-09-26 · Updated 2026-10-15 · System Design
Ultra-Resilient Architecture: Keeping a Service Useful When Parts Fail
Three zones don't help if they all depend on one shared database. Quorums, static stability, multi-region trade-offs, and why replication isn't a backup.
Created 2026-09-21 · Updated 2026-10-15 · System Design
gRPC in Depth: Calling Another Service Without Pretending the Network Is Reliable
Contracts, Protocol Buffers, HTTP/2 streams, deadlines, status codes, retries, load balancing, keepalive, and browsers: how gRPC works and where the sharp edges are.
Created 2026-09-26 · Updated 2026-10-15 · System Design
Algomi Architecture, Part 1: The Backbone Every Product Shares
How Algomi's five web apps, single login, stateless API, shared library, market-hours worker, and hand-applied database migrations fit together.
Created 2026-09-26 · Updated 2026-10-15 · System Design
Algomi Architecture, Part 2: Trading, Insights, Charts, Career, and the AI Guardrails
How a Telegram message becomes a broker order, why recommendations refuse to guess, and how one doorway keeps personal data away from third-party AI models.
Created 2026-09-26 · Updated 2026-10-15 · System Design
Interview Roadmap Appendix: 107 Topics, One Index
A study index for system design and AI system design interviews: every roadmap topic mapped to the post and section that teaches it, the HLD, LLD and AI interview frameworks, a ten-phase study order, and a readiness checklist with links.
Created 2026-09-29 · Updated 2026-10-02 · System Design
Inside a Database: Indexes, Transactions, and Storage Engines
What a database does between your query and the disk: ACID, isolation levels and their anomalies, B-tree and LSM storage engines, write-ahead logs, index design, query plans, connection pools, and how to choose between SQL and NoSQL.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Scaling Data: Replication, Sharding, and Distributed Transactions
When one database is not enough: replication topologies, choosing a shard key, hot partitions, global secondary indexes, and how to keep a business operation correct across services with 2PC, sagas, and the outbox pattern.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Consistency Models, Quorums, and Logical Clocks
What 'consistent' actually means: CAP and PACELC, the ladder from linearizable to eventual, quorum arithmetic, and how vector clocks and CRDTs let replicas agree without a shared clock.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Consensus, Raft, Distributed Locks, and Leader Election
How machines that can crash and messages that can vanish still agree on one answer: the Two Generals problem, Byzantine faults, Raft in plain language, leader election, split brain, and why a lock needs a fencing token.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Caching Deep Dive: Patterns, Invalidation, and Stampedes
Where to cache, which write pattern to use, how to evict, how to invalidate, and how to survive cache stampedes, penetration, and hot keys, with the trade-offs an interviewer expects you to name.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Load Balancing, Global Traffic, TLS, and HTTP Versions
Layer 4 versus layer 7 balancing, algorithms and health checks, DNS and anycast for global traffic, the TLS handshake, what changed between HTTP/1.1, 2 and 3, and where an API gateway fits.
Created 2026-09-29 · Updated 2026-10-15 · System Design
API Design: Pagination, Idempotency, Rate Limiting, and Access Control
REST, GraphQL and gRPC choices, cursor pagination, idempotency keys, versioning, a rate limiter designed end to end (token bucket, sliding window, distributed), and RBAC, ABAC and JWT trade-offs.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Kafka and Messaging Reliability: One Diwali Night in Ravi's Sorting Office
Follow one night in a busy sorting office to learn how Kafka stores and replicates messages, what acks and ISR really guarantee, delivery semantics, rebalancing, lag, poison pills, dead-letter queues, and retries with backoff and jitter.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Data Platforms: Batch, Streaming, CDC, and the Lakehouse
OLTP versus OLAP, batch versus stream, lambda and kappa, windowing and late data, change data capture, data lakes and warehouses, and how to keep a pipeline correct when events repeat or arrive late.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Search, Autocomplete, Web Crawling, Time-Series, and Ad-Click Aggregation
Inverted indexes and ranking, a typeahead service built on tries, a polite web crawler, time-series storage with downsampling, and exactly-once-ish ad-click aggregation.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Location-Based Services: Geohash, Quadtrees, Nearby Search, and Ride Matching
How to answer 'what is near me?' at scale: geohash, quadtrees, S2 and H3 cells, a proximity service, and a ride-hailing dispatcher that ingests millions of moving driver locations.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Collaborative Editing: Operational Transformation, CRDTs, and Presence
How two people type in the same document at once without losing each other's words: operational transformation, CRDTs, the server architecture around them, presence, offline editing, and version history.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Stock Exchange and Broker Design: Order Books, Matching Engines, and Risk
How an exchange matches buyers and sellers deterministically: the limit order book, price-time priority, a single-threaded sequenced matching engine, market data fan-out, and the broker's order gateway and pre-trade risk checks.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Chat and Notification Systems: Real-Time Delivery at Scale
Designing a WhatsApp-style chat (connections, message flow, ordering, delivery receipts, offline sync, group fan-out) and a notification platform (channels, templates, preferences, retries, deduplication).
Created 2026-09-29 · Updated 2026-10-15 · System Design
News Feed and Social Graph: Fan-Out, Ranking, and the Celebrity Problem
Fan-out on write versus on read, the hybrid that tames celebrity accounts, storing the follow graph, feed ranking, counters, and pagination for a timeline that must feel instant.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Storage Case Studies: URL Shortener, Key-Value Store, and File Sync
Three classic designs worked end to end: a URL shortener (ID generation, redirects, analytics), a Dynamo-style distributed key-value store, and a Dropbox-style file sync service with chunking and deduplication.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Capacity Estimation and Load Testing: Back-of-the-Envelope Done Properly
The numbers every engineer should know, a repeatable method for QPS, storage, bandwidth, and server counts, worked examples, and how to check the estimate with load tests and percentile latency.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Production Engineering: Observability, Security, Deployment, Kubernetes, and Cost
What separates a design on paper from a system you can run: metrics, logs and traces, SLOs, zero-trust security, safe deployment strategies, Kubernetes essentials, multi-region cloud choices, and cost control.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Low-Level Design Foundations: OOP, SOLID, and Design Patterns in Java, Told Through a Tea Shop
The object-oriented vocabulary and the handful of design patterns that machine-coding rounds actually use, explained like you are five through one tea shop, then shown in short runnable Java.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Concurrency and Multithreading in Java for LLD: Races, Locks, and Producer-Consumer
One busy restaurant kitchen explains race conditions, locks, deadlock, producer-consumer, thread pools, and thread-safe class design, with runnable Java 17 code and the Java-specific facts (memory model, volatile, ReentrantLock, ConcurrentHashMap, virtual threads) interviewers probe.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Machine Coding Case Studies, Part 1: Parking Lot, LRU Cache, Rate Limiter, Vending Machine (Java, via a Shopping Mall)
Four classic machine-coding rounds solved in modern Java through one neighbourhood shopping mall: the basement parking lot, the front-shelf LRU cache, the food-court token bucket, and the snack vending machine, each with design reasoning and follow-ups.
Created 2026-09-29 · Updated 2026-10-15 · System Design
Machine Coding Case Studies in Java, Part 2: A Weekend Hotel Trip with Splitwise, Elevator, Tic-Tac-Toe, Logger, and Design Sketches
One weekend hotel trip, told in modern Java: expense sharing with debt simplification, a lift scheduler, tic-tac-toe with O(1) win checks, a pluggable trip diary (logger), plus class sketches for seat booking, notifications, a library, cab dispatch, food delivery, and chess.
Created 2026-09-29 · Updated 2026-10-15 · System Design
ML System Design and Recommendation Systems: From Data to Ranking in Production
The end-to-end shape of a production ML system (framing, data, features, training, serving, monitoring) and the two-stage recommender pattern with candidate retrieval, ranking, metrics like NDCG, and cold start.
Created 2026-09-29 · Updated 2026-10-15 · System Design
LLM Serving Infrastructure and Capacity Planning: SLOs, Parallelism, and Autoscaling
Turning the mechanics of LLM inference into a capacity plan: TTFT and tokens-per-second targets, how many GPUs a model needs, tensor and pipeline parallelism, prefix caching, speculative decoding, serving engines, streaming, and autoscaling.
Created 2026-09-29 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
Vector Search and Production RAG: ANN Indexes, Multi-Tenancy, Freshness, and Evaluation
How approximate nearest neighbour indexes (HNSW, IVF, product quantisation) work and trade recall for speed, then the production concerns the basic RAG guide skips: permission-aware retrieval, multi-tenancy, real-time updates, hybrid search, and RAG evaluation.
Created 2026-09-29 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
Agent Patterns: State Machines, Memory, and Tool Safety
The design patterns behind reliable LLM agents: ReAct and plan-and-execute loops, agents as explicit state machines, the memory taxonomy, multi-agent trade-offs, and the safety rules for tools that change the world.
Created 2026-09-29 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
AI Security, Reliability, and Cost Control: Prompt Injection, Failover, and Budgets
Defending LLM applications against prompt injection and token exhaustion, keeping them up with provider failover, circuit breakers and graceful degradation, and keeping the bill predictable with budgets, routing, and caching.
Created 2026-09-29 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
AI System Design Case Studies and the Interview Framework
An eight-step framework for AI system design interviews, applied to eleven case studies: enterprise ChatGPT, coding assistant, AI search, knowledge assistant, support AI, document search, AI gateway, multi-model platform, agent platform, and evaluation and observability platforms.
Created 2026-09-29 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
Harness Engineering: Making AI Agents Repeatable
Prompts tell an agent what to do; a harness decides what it knows, what it may touch, and what counts as done. Five subsystems, the session lifecycle, and a working Java shift-handover model, told as a hospital night ward.
Created 2026-10-02 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
Agent Loops, Maker-Checker, Skills and Tool Gateways
How to turn repeated prompting into a controlled loop: goal loops, a separate checker, retry budgets, state kept outside the loop, reusable skills, and a tool gateway with permissions, timeouts and idempotency. Told as a child doing homework.
Created 2026-10-02 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
Agent Graphs, Isolation, Rollback and Harness Evaluation
When one loop is not enough: model the workflow as a graph, fan work out and back in, isolate parallel agents, roll back on failure, gate risky steps on a human, and measure the harness with ablation tests. Told as building a house.
Created 2026-10-02 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
How Kafka Log Compaction Keeps Only the Latest Value per Key (and Where It Bites You)
What log compaction actually does inside a Kafka broker: dirty and clean sections, the cleaner's offset map, tombstones and delete.retention.ms, the settings that decide when a record really disappears, and the three interview questions that trip people up.
Created 2026-10-15 · Updated 2026-10-15 · System Design
API Monitoring: Proving the Library Can Still Reserve a Book
A health check that passes while nobody can reserve a book: how to monitor API outcomes with synthetic journeys and real-request telemetry, keep test traffic safe, turn failures into actionable alerts, and defend the cost, privacy and recovery decisions in an interview.
Created 2026-10-17 · Updated 2026-10-17 · System Design
Enterprise MCP Gateway: One Front Door for Thousands of AI Agent Tools
Why a company with hundreds of MCP servers needs a gateway, and how a control plane, a data plane, tool-level authorization, progressive discovery and response projection work, using Uber's published MCP Gateway as a case study.
Created 2026-10-04 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
Multi-Tenant RAG: Retrieving the Right Answer from the Right Customer
How to stop one company's documents leaking into another company's answers: tenant isolation at ingestion, retrieval, context assembly and caching, and how to choose between a shared index, namespaces and dedicated indexes.
Created 2026-10-04 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
Agent Architecture: Designing AI Systems That Take Actions, Not Just Answers
The agent loop, designing tools the model can actually use correctly, reactive versus planned execution, memory and compaction, multi-agent orchestration, and the guardrails that keep an acting AI safe.
Created 2026-09-18 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
LLM Serving and Inference Cost: What Actually Determines Latency and Price
Why a demo that works for one user can collapse at scale — prefill vs. decode, the KV cache, continuous batching, PagedAttention, quantization, caching, and a worked cost estimate.
Created 2026-09-18 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
RAG System Design: How to Ground an AI's Answers in Your Own Data
Why a fluent AI answer can still be wrong — chunking, embeddings, hybrid search and reranking, and when retrieval beats fine-tuning or a bigger context window.
Created 2026-09-18 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering
AI Evals and Observability: How to Check Whether an AI Actually Works
Why a fluent AI answer isn't proof of success — evals, traces, groundedness, graders, and the release checks that catch failures a good average score would hide.
Created 2026-09-16 · Updated 2026-10-15 · AI Solutions Architecture & AI Engineering