Projects
Sanitized architecture case studies — what each system is, how it is wired, and why the pieces are there.
01
Crypto Tracker - Local Data Pipeline
An end-to-end local data pipeline: crypto and FX data ingested with dlt, warehoused in DuckDB, modelled with dbt into a medallion + star schema, orchestrated by Airflow 3.
dlt ingests CoinGecko and Frankfurter APIs into a DuckDB landing zone; dbt builds bronze trust-filter views, silver cleaned staging, and a gold star schema (dims, facts, marts). One Airflow 3 DAG runs ingest-then-build daily through a single-slot pool that serializes DuckDB writes. No API keys, no cloud account.
- data-engineering
- python
- SQL
- Airflow
- dbt
- DuckDB
- dlt
View repository ↗02
Blanche - Online Clothing E-commerce Platform
Full stack e-commerce website written in React Typescript + Gin Golang. Enabled caching mechanism through Redis.
React TypeScript storefront over a Gin (Go) REST API backed by PostgreSQL. A Redis read-through cache serves repeat catalog queries, and authentication uses JWT with role-based access.
- web-development
- frontend
- React
- backend
- Go
- database
- Redis
View repository ↗03
Sushee - Online Restaurant App
An Online Restaurant built using Golang, PostgreSQL, and React.
Three-tier ordering app: React SPA -> Go REST API -> PostgreSQL, run locally with Docker Compose. Orders flow through an explicit state machine from placement to kitchen fulfillment.
- web-development
- frontend
- React
- backend
- Go
- database
View repository ↗04
E-Wallet Backend Service
An E-Wallet API built using Golang and PostgreSQL.
Transactional ledger service in Go on PostgreSQL: double-entry balance records, idempotency keys on transfer endpoints, and row-level locking so concurrent transfers cannot lose updates.
- software-development
- backend
- Go
- database
View repository ↗05
MasTrendy - Daily Update Discord Bot
A Discord bot that publishes a curated daily trends digest.
Scheduled Python pipeline fetches daily trend sources, normalizes and deduplicates them, persists the results, formats a digest, and pushes it to Discord channels via webhook on a cron schedule.
- software-development
- bot
- data-engineering
- python
View repository ↗06
Hotel Booking Cancellation Prediction Model
Predictive model categorizes bookings as likely or unlikely to be canceled, achieving an accuracy of 83.56%.
Feature-engineered classification pipeline (scikit-learn) predicting hotel booking cancellations at 83.56% accuracy; the trained artifact is persisted and reused for batch scoring.
- data-science
- data-analytics
- machine-learning
View repository ↗