Build real systems
Theory is easy. Employers want builders. Each project walks you through requirements → design → implementation → scaling - exactly how it works at a real company.
What's inside every project
Requirements
Real functional + non-functional specs - just like a product team hands to engineers.
API Design
Contract-first REST API definitions with endpoints, request & response schemas.
DB Schema
Normalized schemas with justifications for every design decision made.
Scaling Plan
How to evolve from 100 to 10M users - sharding, caching, CDN, and beyond.
Hands-on learning
Pick a project and start building
ChurnCast: End-to-End Data Science Case Study on Subscription Churn
TiffinBox, a meal-subscription startup, loses about 9 % of its paying customers every month, and its retention team can call only 10 % of them with an offer. Over nine phases you build the churn model they need, like a professional data science team: frame the decision and the label, write a data contract, a timeline and a modelling plan with ADRs, then build the pipeline test-first in your track: load four deliberately messy CSVs as text, apply documented cleaning rules that count every row they drop, check the contract, build point-in-time features that only use what was known at each month-end snapshot (and avoid the leaky cancellation_reason column), compare two baselines (the training churn rate and the team's five-flag rule) with two logistic regressions using grouped 5-fold cross-validation, then evaluate the champion once on out-of-time test months: AUC, average precision, Brier, calibration (ECE), precision and recall of the top-10 % call list, net value per month, slices, an age-band fairness check and a leakage scan, all behind a release gate. Golden files, invariants, leakage and contract-violation tests pin every number, and the Python, SQL and R tracks must agree (features to 1e-9, metrics and coefficients to 1e-6). CI executes the notebook or R Markdown report, runs the gate and checks parity; a generated model card and a FastAPI prediction API with an optional AI explainer (numbered facts in, guardrails out, a deterministic template as fallback) ship in Docker as v1.0.0. Then the model meets the real world: a new acquisition channel breaks the contract, prices change, and the deployed model's calibration slips. You version the contract, add drift (PSI) and realised-performance monitoring, recalibrate and retrain a challenger, decide champion vs challenger on a holdout month, add request ids, JSON logs and Prometheus metrics to the API, and write a blameless post-mortem (v1.1.0). Pick Python (pandas, scikit-learn, Jupyter), SQL (DuckDB) or R (tidymodels, R Markdown); the data, contract, golden files and API are shared. Everything runs on a laptop with no paid service and no API key.
DevBlog: Secure Blog REST API with JWT, Roles and AI Moderation
Over nine phases you build DevBlog like a professional team: requirements, design, a test-first implementation of JWT auth with refresh-token rotation, role-based access, posts with optimistic locking, keyset pagination and full-text search, comments with AI moderation and a human in the loop, CI/CD with GitHub Actions, Docker, optional free-tier deployment, observability and a retrospective. Pick Java, Python or TypeScript; the spec, database and tests are shared. Everything runs on your laptop, and the AI features work offline.
FunnelLens: Data Analyst Case Study on a Messy E-commerce Funnel
Mobile is about 69 % of FunnelLens's sessions but only about 61 % of its revenue, and the Head of Product wants a quarter of engineering time to rebuild the mobile checkout. Over nine phases you find out what the data supports, like a professional analytics team: turn the request into answerable questions with acceptance criteria, write a data contract and an analysis plan with ADRs, then build the analysis test-first in your track: load four deliberately messy CSVs as text, apply documented cleaning rules that count every row they drop, check the contract, build the funnel by device, channel and week, and test the hypothesis three ways (a two-proportion z test, a Cochran-Mantel-Haenszel test that adjusts for acquisition channel and exposes a Simpson's paradox, and Welch's t test on order values) before sizing the opportunity against the rebuild threshold. Golden answers, invariants and contract-violation tests pin every number, and the Python, SQL and R tracks must agree to 1e-9. CI executes the notebook or R Markdown report and checks parity; a Streamlit dashboard with an optional AI narrator (numbered facts in, guardrails out, a deterministic template as fallback) ships in Docker. Then a data refresh breaks the pipeline on purpose: a new acquisition channel and a broken tracking week lead to a contract version bump, monitoring, a post-mortem and v1.1. Pick Python (pandas, Jupyter), SQL (DuckDB) or R (tidyverse, R Markdown); the data, contract, golden answers and dashboard are shared. Everything runs on a laptop with no paid service and no API key.
Pipewright: End-to-End Data Pipeline for a CEO Dashboard
BazaarBox, a direct-to-consumer home & kitchen brand, wants one dashboard that is correct by 06:30 IST every day: revenue by category and region, CAC by channel, Prime membership MRR, the top 10 products of the last 30 days and 30-day cohort retention. Over nine phases you build the pipeline behind it like a data engineering team: write requirements with SLAs and acceptance criteria, design the architecture, the bronze/silver/gold layers, a data contract and five ADRs (including why Airbyte, a Spark cluster and Metabase were swapped for lighter tools), then build it test-first. A seeded, offline copy of BazaarBox's systems (an orders and customers API with bearer tokens, cursor pagination, updated_at windows and injected 429/503 faults, plus daily payment, ad-spend and catalogue file drops) comes with real-world defects: mixed time zones, re-delivered orders, duplicate payment events, rupee amounts as text, city spellings, test orders and PII. You write idempotent ingestion into immutable lake partitions, silver and gold models in dbt (SQL track) or PySpark (Python track) whose gold tables must equal the golden files and each other row by row, declarative data-quality checks with severities and a publish gate built on an atomic swap, dbt data and unit tests, and a GitHub Actions pipeline that runs dbt build on a fixture and both tracks on the full history. v1.0.0 adds an Airflow 3 schedule, a Streamlit dashboard, an optional AI incident explainer with guardrails and an offline fallback, and a hardened Docker Compose stack under 6 GB of RAM. Then the pipeline meets its second month: a new channel and a new order status block publishing, coupons change the schema, and a partner file arrives late. You version the contract additively, make the facts incremental and prove them equal to a full refresh, add backfills, freshness, volume and SLA monitoring with alerts, prove backward compatibility with v1.0, and write a blameless post-mortem (v1.1.0). Everything runs on a laptop with no paid service and no API key.
RateGuard: Distributed Rate Limiter with Redis, Lua and an AI Policy Advisor
Over nine phases you build RateGuard like a professional team: requirements and threats, design with ADRs, a test-first implementation of a token bucket and a sliding window counter as pure functions and as atomic Redis Lua scripts that use one clock, a Redis Cluster-ready key layout with hash tags, tenants with SHA-256 hashed API keys and a constant-time admin token, the standard rate-limit headers and Retry-After, per-minute usage statistics and an AI policy advisor with an offline fallback, integration tests that race parallel requests for the last token, CI/CD with GitHub Actions, Docker, optional free-tier deployment, a game day that freezes Redis, a circuit breaker with per-tenant fail modes, observability and a retrospective. Pick Java, Python or TypeScript; the spec, the Lua scripts and the tests are shared. Everything runs on your laptop, and the AI feature works offline.
ShopFlow: E-commerce Backend with Idempotent Checkout, Payments and AI Search
Over nine phases you build ShopFlow like a professional team: requirements and threats, design with ADRs, a test-first implementation of a product catalog with full-text search, a server-side cart with coupons, money as integer paise, checkout with an Idempotency-Key and a guarded stock decrement that cannot oversell, orders with a state machine, HMAC-signed and de-duplicated payment webhooks, a transactional outbox, AI search assist and product description drafts with offline fallbacks, CI/CD with GitHub Actions, Docker, optional free-tier deployment, observability, a payment-window migration and a retrospective. Pick Java, Python or TypeScript; the spec, database and tests are shared. Everything runs on your laptop, and the AI features work offline.
SnapLink: URL Shortener with Analytics and AI Insights
Over nine phases you build SnapLink like a professional team: requirements, design, a test-first implementation, CI/CD with GitHub Actions, Docker, optional free-tier deployment, observability and a retrospective. Pick Java, Python or TypeScript; the spec, database and tests are shared. Everything runs on your laptop, and the AI feature works offline.
StudyBuddy: RAG Study Assistant with Cited, Evaluated Answers
Over nine phases you build StudyBuddy like a professional team: requirements, threats (prompt injection, cross-tenant leaks, invented citations) and user stories, a design with ADRs, then a test-first implementation of a complete retrieval-augmented generation pipeline: PDF and text ingestion with page-aware chunking, deterministic offline embeddings pinned by golden fixtures, hybrid retrieval in PostgreSQL (pgvector cosine search plus full-text search fused with reciprocal rank fusion), coverage-based abstention, extractive answers with page citations, an optional LLM whose answers are validated like user input, and streaming over server-sent events. Workspaces isolate tenants with SHA-256 hashed API keys. An evaluation harness measures recall, MRR and citation precision on every change, CI runs it in Docker, and a game day that makes the LLM hang leads to circuit breakers, degraded retrieval, daily quotas, metrics and a post-mortem. Pick Java, Python or TypeScript; the specs, golden files, tests, evaluation set and the chat UI are shared. Everything runs on your laptop with no API key; Ollama or a free Gemini key are optional.