About this project

DE Portfolio — a real data platform

This portfolio is not mockups — it is a live homelab running a production-grade data engineering stack. Every chart, table and metric on this site comes from actual pipelines processing real financial and market data.

How the pipeline works

1

Ingest

APIs polled by Airflow DAGs every 15–60 min. Binance WebSocket streams to Kafka in real time.

2

Stream

Kafka topics buffer raw trade data. ClickHouse consumer writes 25M+ rows directly from broker.

3

Transform

dbt models build staging → mart layers in Neon PostgreSQL. Idempotent, tested, documented.

4

Visualize

Grafana reads ClickHouse for low-latency metrics. Superset queries PostgreSQL for BI reports.

Tech Stack

Homelab

OSUbuntu 24.04 LTS
CPUIntel Core i5
RAM16 GB DDR4
Storage500 GB SSD
NetworkTailscale VPN mesh

Data Services

OrchestrationApache Airflow 2.9
StreamingApache Kafka
OLAPClickHouse
OLTPPostgreSQL 16 (Neon)
Transformdbt Core

BI & Monitoring

DashboardsGrafana
BI AnalyticsApache Superset
CI/CDGitHub Actions
ContainersDocker / Docker Compose

Data Sources

Crypto pricesCoinGecko REST API
Forex ratesNBK Kazakhstan API
WeatherOpen-Meteo API
NewsTengrinews scraping
TradesBinance WebSocket (25M+ rows)

Live services

PostgreSQL Schema

staging schema

staging.stg_crypto_prices

coin_id, symbol, current_price_usd, price_change_pct_24h, market_cap_usd

staging.stg_forex_rates

currency_code, rate_kzt, units, rate_date

staging.stg_news

title, category, published_at, url

mart schema

mart.mart_daily_crypto_kzt

trade_date, symbol, avg_price_usd, avg_price_kzt

mart.mart_forex_trend

rate_date, currency_code, rate_per_unit, day_change