Spring Boot + Angular revenue analytics with a Python ML service for 30-day forecasting, anomaly detection, cancellation risk and rate recommendations.

Spring Boot 3 REST API with JPA/Hibernate on MySQL, RFC 7807 errors, Swagger docs and 29 JUnit/MockMvc tests. A FastAPI service blends Holt-Winters with a booking-pickup model (2.6% MAPE in a 28-day backtest vs 13.5% for the naive baseline), flags unusual bookings with Isolation Forest and control rules, and scores cancellation risk with a Random Forest.

Flask app with passwordless email OTP login, an NLP + Random Forest symptom checker, appointment booking and a real-time queue over Server-Sent Events.

Rebuild of my E-Hospital minor project. OTP codes are HMAC-hashed with expiry, attempt limits and rate limiting; CSRF protection and role-based access throughout. Free-text symptoms are parsed with negation handling and typo correction; the classifier reaches 99% top-3 accuracy on held-out synthetic cases. 33 pytest tests.

Next.js ticketing app with bcrypt passwords, signed session cookies, server-side role checks and transactional seat-capacity control.

MySQL schema with foreign keys and constraints; bookings lock the event row in a transaction so the last tickets can’t be sold twice. Per-slot capacity for stadium tours, fan and admin dashboards with live revenue figures, and Vitest tests for session forgery, validation and capacity.

Next.js + MongoDB storefront with search, brand filters, cart and server-side checkout that reserves stock atomically.

Orders are priced on the server from the catalogue, and stock is reserved with conditional atomic updates so two shoppers can’t buy the last unit. Store management needs an admin key; search input is regex-escaped and product updates are field-whitelisted.

Interactive simulator that visualises cache hits, misses and LRU/FIFO replacement for a stream of memory addresses.

Runs entirely in the browser with React and TypeScript: enter addresses, pick a replacement policy and watch the cache lines, hit rate and evictions update step by step.

Hour-by-hour electricity forecasting with leak-free lag features and a chronological comparison of Ridge, Random Forest and Gradient Boosting.

Loads and cleans the UCI household power dataset (2M minute-level rows) into hourly kWh, engineers calendar, lag and rolling features, and picks the best model on a validation split. On held-out data it cuts error by about 27% versus a same-hour-last-week baseline. CLI, FastAPI endpoint and pytest suite.