Raw in. Value out.

Aleksa Milic

I build and run data platforms end to end: Spark and Azure Data Factory pipelines, warehouses that stay fast at billions of rows, and the CI/CD and infrastructure under them.

02 / Selected work

Things I’ve built

  1. 012026

    Niš Urban Development Radar

    Construction activity in Niš from official Statistical Office open data: normalised and validated in Python, loaded into PostgreSQL, served as an Apache Superset dashboard.

    Role
    Ingestion and validation, the SQL schema and views, the dashboard as code, and the Dockerised deploy on one VM.
    Outcome
    Reproducible from a clean clone. Sources are checksummed and gaps are never zero-filled. The panel below is a self-contained rebuild.
    • Python
    • PostgreSQL
    • Apache Superset
    • Docker
    • Caddy

    Live dashboard

    Pipeline and methodology on GitHub

  2. 022026

    When do you actually need a cluster?

    A reproducible TPC-H benchmark: Apache Spark (local, plus a real 2-worker cluster) against DuckDB, Polars and DataFusion on identical Parquet. How far does one machine get before a cluster earns its overhead?

    Role
    Engine adapters, the timing and memory harness, a cross-run correctness gate, the Dockerised cluster, CI, and the report generator.
    Outcome
    The in-process engines finish ~10x faster than Spark at SF1 and SF10, and the gap holds as data grows. Every run gated on 96 correctness checks.
    • Apache Spark
    • DuckDB
    • Polars
    • DataFusion
    • Docker
    • TPC-H
    Source on GitHub
    TPC-H workload time, log scale: in-process engines versus Spark, SF1 to SF10
  3. 032026

    Harvest Yield Warehouse

    A dimensional warehouse for a produce cooperative’s harvest deliveries: an operational feed modelled into a Kimball star schema with slowly-changing dimensions and a tested dbt layer.

    Role
    The data model, the dbt project, the test suite, CI, and the published lineage site. A rebuild of an old Oracle and Apache NiFi coursework mart.
    Outcome
    Rebuilds from a clean clone in one command. Every CI run is gated on 65 data tests. Model docs and the lineage graph deploy to GitHub Pages on each push.
    • dbt
    • DuckDB
    • Kimball / SCD2
    • GitHub Actions
    • SQL
    Source on GitHub

    Lineage graph

    11 models5 seeds65 tests

    30 not_null13 unique9 accepted_values8 relationships5 custom

    Open the full dbt docs

03 / Skills & experience

What I work with

I’m a data engineer with four years building pipelines and cloud data platforms across AWS, Azure, and GCP, mostly on Databricks and Azure Data Factory. Enough of my time has gone into DevOps (containers, CI/CD, infrastructure-as-code) that I can take a pipeline from ingestion to production without handing it off.

Ingestion & orchestration

  • Azure Data Factory
  • Apache Airflow
  • AWS Glue
  • AWS DMS
  • Python

Processing

  • Apache Spark
  • PySpark
  • Databricks
  • Microsoft Fabric

Storage & modelling

  • Delta Lake
  • Medallion architecture
  • AWS Redshift
  • Azure Synapse
  • BigQuery
  • Advanced SQL

Platform, DevOps & BI

  • Docker
  • Kubernetes
  • Terraform
  • Multi-cloud CI/CD
  • Power BI
  1. Jan 2026 – Present

    Data Engineer

    Ingsoftware / ASML

    • Run distributed processing on Databricks and Spark over 10+ TB, with Delta Lake for ACID transactions on AWS. Extended the same patterns to GCP (BigQuery, Dataflow) for a multi-cloud setup.
    • Cut average runtime on critical Spark jobs by 40% through partition tuning, caching, and cluster right-sizing.
    • Own data governance across Databricks and Delta Lake: quality checks, access control, and lineage for production datasets.
  2. Jun 2024 – Jan 2026

    Data Engineer

    Vega IT

    • Built and ran 20+ ETL/ELT pipelines on Azure Data Factory and Airflow, with reconciliation and data-quality controls gating every load.
    • Architected warehouses on AWS Redshift and Databricks Delta Lake over 5+ billion records: partitioning, distribution keys, and incremental models sized to the query patterns.
    • Shipped 25+ Power BI dashboards and ran the platform DevOps: ECS/EKS with Docker and Kubernetes, Terraform IaC, multi-cloud CI/CD.
  3. Jul 2022 – Jun 2024

    Data Analyst

    Gemini Software

    • Automated transformation and ingestion workflows in Python (Pandas, NumPy), replacing steps that had been run by hand.
    • Handled cleansing and collation from internal and external sources. Wrote complex SQL with advanced joins, window functions, and CTEs.
    • Integrated third-party APIs and contributed to AWS/Azure cloud-migration work.

Education

  • 2023 – 2025

    MSc, Data Science and Engineering

    Faculty of Electronic Engineering, Niš

    GPA 9.7 / 10. Thesis: optimising distributed data-pipeline architectures on cloud platforms.

  • 2019 – 2023

    BSc, Computer Science and Informatics

    Faculty of Electronic Engineering, Niš

    GPA 8.7 / 10. Thesis: benchmarking distributed data-processing frameworks for large-scale analytics.

Certifications

04 / Contact

Get in touch

I’m at Ingsoftware these days, on ASML’s data platform. Open to work on data platforms, Spark performance, or multi-cloud. Email is the quickest way to reach me.

Download CVPDF
GitHub LinkedIn

© 2026 Aleksa Milic