Production Technician
Iritel Jun 2025 – May 2026- Assembled and soldered power-electronics components for industrial equipment while maintaining strict quality-control standards.
★ Belgrade, Serbia · Open to the trail ahead ★
Luka Pokrajac
I build real-world data pipelines — from embedded sensors to streaming ingestion, ETL orchestration and live dashboards.
01 About
Electrical-technician background turned data & software engineer. I’ve gone from programming 6-axis CNC machines and soldering power electronics to designing containerized, production-style data pipelines. I like systems where the data is real — coming off actual hardware — and I care about the unglamorous parts: idempotent loads, decoupled ingestion, and one-command setups that just work.
Languages
- Python
- C / C++
- SQL
- Bash
Data & Infra
- Kafka
- Airflow
- PostgreSQL
- Spark
- Docker
Cloud & Tools
- AWS (EC2, RDS)
- Grafana
- Streamlit
- Git
Hardware
- ESP32 / ESPHome
- PlatformIO
- MQTT
- Sensors
02 Experience
Maintenance Technician
Mei-Ta Aug 2023 – Apr 2024- Programmed, configured, and operated 6-axis CNC grinding machines for high-volume production of precision cutting tools.
- Optimized machining parameters and process settings to improve tool life, dimensional accuracy, and production throughput.
- Worked within an industrial manufacturing environment using automated production equipment and quality-control procedures.
03 Projects
IoT Telemetry Pipeline
↗Real-hardware telemetry for a mushroom fruiting chamber: an ESP32-C6 publishes 7 environmental sensors over MQTT into a decoupled, containerized stack.
- Normalized heterogeneous sensor topics into one canonical event schema.
- Decoupled ingestion from storage with a Kafka stage so the producer never blocks on the DB.
- Containerized Kafka (KRaft) + Grafana with an auto-provisioned PostgreSQL source for one-command setup.
- Custom PlatformIO/C++ firmware for an ESP32-CAM, converting RGB565 frames to JPEG in software.
Wear Index ETL Pipeline
↗Fully automated hourly ETL that ingests live Open-Meteo weather data and computes a custom “feels-like” index, deployed to AWS.
Live demo → wear.lukapokrajac.com
- Custom index from rolling-window averages and wind-chill logic.
- Idempotent incremental loads with 3-hour overlap and PostgreSQL UPSERT — zero duplication.
- Orchestrated with hourly Airflow DAGs; auto-refreshed materialized views for instant querying.
- Two Docker Compose stacks on a shared network; one-command setup via Makefile.
- Production instance on AWS EC2 with persistent PostgreSQL on RDS.
Real-Time Telemetry Streaming
↗Mission-control-style streaming pipeline ingesting simulated spacecraft telemetry at 1 event/sec through Python → Kafka → Postgres → Streamlit.
- Python producer/consumer over Kafka with validation, cleaning and rule-based anomaly detection.
- Dual-storage model (raw + cleaned) in PostgreSQL.
- Live Streamlit dashboard for time-series visibility.
- Spark window-aggregation batch job; containerized with Docker Compose + Makefile.
E-commerce Analysis
↗Exploratory analysis of the Olist Brazilian e-commerce dataset, turning messy relational data into actionable business insight.
- Merged and cleaned multiple relational tables with pandas; resolved missing values and inconsistent timestamps.
- Delivery-performance analysis comparing estimated vs. actual dates to find logistics bottlenecks.
- matplotlib visualizations of price distributions and delivery-delay patterns.
04 Education
University of Belgrade — School of Electrical Engineering (ETF)
2024 – 2025Technical School Obrenovac — Electrical Technician for Process Control
2019 – 202305 Get in touch
Open to data engineering and backend roles. The fastest way to reach me is email.
lukapokrajac04@gmail.com