Skip to content
Ömer Faruk Ballı

Senior Data EngineerAI/ML Engineer

Ömer FarukBallı

Raw signal in. Working systems out.

I build production GenAI systems and enterprise data platforms. Seven years, nine teams — and every engagement closed with a working system, documented and handed over.

Send a messageDownload résumé

Open to long-term in-house roles

02Origin

Trained as an economist.Working as an engineer.

I studied economics at Istanbul Technical University and wrote my graduation project on what drives electricity prices in Turkey. A master's in data science at Galatasaray University followed, with a thesis on graph networks.

Since 2019 I have worked in factories, quick-commerce, mobile games, music and global enterprise programs — first as the data scientist building the models, then as the engineer building the platform they run on. Most of that work was fixed-scope: build, deliver, hand over.

Now I am looking for the opposite: a long-term in-house role, where I stay with what I build and grow it.

years shipping data & AI to production
7+
zero-downtime platform migrations
2
countries led on a global SAP S/4HANA program
3
data connectors shipped in AutoData
30+

03Pipelines

Nine teams.Every one left with a working system.

  1. 01 / 09Apr – Sep 2026

    Air Liquide

    Data Excellence Lead · Contract

    Led data strategy across Turkey, Bulgaria and Romania for Align, the global SAP S/4HANA implementation.

    • Translated business requirements into master-data specifications across manufacturing and retail operations.
    • Architected master-data preparation across ERP and satellite systems, with data-quality governance through SIT, UAT and cutover.
    • Closed with a structured handover of the governance framework, cutover documentation and executive analytics.
    • SAP S/4HANA
    • Master data
    • SQL
    • ETL
    • Governance

  2. 02 / 09Dec 2025 – Apr 2026

    BGTS

    Senior Data Scientist · Contract

    Built and operated AutoData, a full-stack ML data-preparation platform.

    • Six-stage pipeline, 30+ data connectors and a published PyPI SDK with REST API.
    • Led a zero-downtime production migration via live public-IP cutover (31 GB → 48-core / 992 GB VM), eliminating OOM failures.
    • Hardened the platform with prompt-injection guards, read-only SQL validation and Fernet credential encryption.
    • Next.js
    • Flask
    • PostgreSQL
    • Redis
    • Azure
    • H100 GPU

  3. 03 / 09Jan – Nov 2025

    Philip Morris International

    Senior Data Engineer · via Adesso · Contract

    Led the migration from legacy AWS (MS-SQL, SSIS) to Snowflake and Matillion with zero downtime.

    • Architected the full warehouse in Snowflake: SCD Type 2 dimensional models across stage, integration and presentation layers.
    • Led a team of three junior data engineers — task planning, code reviews and mentoring.
    • Handed the platform and its Power BI dashboards over to the internal team at contract end.
    • Snowflake
    • Matillion
    • PySpark
    • Power BI
    • AWS

  4. 04 / 09Aug 2024 – Jan 2025

    Multi

    AI/ML Engineer · Contract

    Architected a production enterprise LLM assistant on open models, chosen for data privacy.

    • Multi-model orchestration across LLaMA, Gemma and Mistral, with a RAG pipeline on ChromaDB.
    • Secure multi-tenant architecture: session management, per-tenant document repositories, prompt versioning.
    • WhatsApp integration, REST APIs and multi-format document processing — live at handover.
    • LLaMA
    • Gemma
    • Mistral
    • ChromaDB
    • RAG

  5. 05 / 09Jul 2023 – Aug 2024

    Warner Music Group

    Data & AI Engineer · via BGTS · Contract

    Sole data & AI engineer for Songkick, reporting directly to the CTO.

    • Developed a GenAI content-safety system combining Azure Content Safety with custom LLM workflows.
    • Led the end-to-end data migration for the Songkick app: pipelines, dimensional models, catalog and documentation.
    • Redesigned the Looker ecosystem on Snowflake with clustering-based user segmentation for high-profile artists including Dua Lipa.
    • Azure Content Safety
    • Snowflake
    • Looker
    • LLM workflows

  6. 06 / 09Dec 2022 – Jul 2023

    SmartIQ

    AI/ML Engineer · Contract

    Engineered a hybrid LLM and classical-ML pricing system for vehicle listings.

    • Built multi-label BERT text classification for hierarchical listing attributes.
    • Architected Airflow orchestration on GCP from scratch, with automated scheduling and failure recovery.
    • Applied model quantization for mobile deployment.
    • XGBoost
    • BERT
    • Airflow
    • GCP

  7. 07 / 09Jul 2021 – Jul 2022

    Rollic Games

    Data Engineer & Analytics · via Zynga

    Built real-time player segmentation serving millions of daily active users.

    • Engineered ETL pipelines from game servers and third-party APIs into Snowflake.
    • Delivered a Looker suite for monetization, retention and game-economy analytics.
    • Ran weekly A/B tests and churn analysis that drove product decisions on retention.
    • Snowflake
    • Looker
    • A/B testing
    • Segmentation

  8. 08 / 09Jul 2020 – Jul 2021

    Getir

    Data Scientist

    Engineered a real-time recommendation engine and churn models serving millions of users.

    • Took ML projects end to end: dimensional modeling, SQL transformations, models deployed via Airflow on AWS.
    • Built automated campaign systems that increased basket size and retention.
    • Collaborative filtering
    • Airflow
    • AWS
    • SQL

  9. 09 / 09Aug 2019 – Aug 2020

    Groupe Renault

    Data Scientist

    Architected a predictive-maintenance platform on GCP with real-time FANUC robot telemetry.

    • Built anomaly detection on torque and vibration signals to minimize unplanned downtime.
    • Built energy forecasting across 11 manufacturing departments, benchmarking five time-series models.
    • GCP
    • Prophet
    • LSTM
    • XGBoost
    • Anomaly detection

04Graph

From source systemto the model that uses it.

I am most useful where these meet: an AI system shipped together with the pipelines, evaluation and monitoring it depends on.

  • 01

    GenAI & LLMs

    Applications that run in production

    • RAG & GraphRAG
    • Knowledge graphs (Neo4j, Cypher)
    • LangChain, LangGraph
    • Agentic AI
    • Multi-model orchestration
    • Vector databases
    • Hallucination control
    • LLM security

  • 02

    ML & MLOps

    From analysis to a served model

    • XGBoost, LightGBM
    • PyTorch, TensorFlow
    • Transformers & BERT
    • Time series
    • Recommendation systems
    • Anomaly detection
    • MLflow, drift monitoring
    • A/B testing

  • 03

    Data platforms

    Warehouses people can trust

    • Snowflake
    • SCD Type 2 modeling
    • BigQuery
    • Redshift
    • Databricks
    • PostgreSQL
    • MongoDB
    • Redis

  • 04

    Data engineering

    Pipelines that survive migration

    • Airflow
    • Matillion, dbt
    • Kafka, PySpark
    • ETL / ELT
    • Medallion architecture
    • Data quality & governance
    • Docker, Kubernetes

  • 05

    Enterprise systems

    Master data at program scale

    • SAP S/4HANA master data
    • Salesforce
    • Workday
    • Concur & Wise
    • SIT / UAT
    • Cutover management

  • 06

    Cloud & BI

    AWS, GCP and Azure in production

    • AWS (EMR, Glue, SageMaker)
    • GCP (Vertex AI)
    • Azure (ML, Content Safety)
    • Looker
    • Power BI
    • Tableau

  • 07

    Mobile & product

    Two consumer apps, shipped solo

    • React Native (Expo)
    • Skia, Reanimated
    • Native modules (C/C++, JNI)
    • Supabase
    • App Store & TestFlight

  • 08

    Languages

    The ones I reach for

    • Python
    • SQL (Snowflake, T-SQL)
    • TypeScript / JavaScript
    • C / C++

05Constellations

Things I builtend to end.

Products and platforms where I owned the whole path, from the first commit to the people using it.

06Research

Where the methodcomes from.

  • M.Sc. in Data Science2021 – 2022

    Galatasaray University

    Machine learning and graph-network analysis of Uber and Lyft ride data

    Question
    Do features extracted from the ride network itself improve a price model?
    Method
    Built a graph of pick-up and drop-off points with NetworkX, derived hub and authority scores for every stop, and compared linear regression and decision trees with and without them.
    Finding
    Across 637,976 rides and 88 features, the graph features gave both models a small but consistent gain in explanatory power.
    • Graph networks
    • NetworkX
    • Regression
    • Decision trees

  • B.Sc. in Economics2015 – 2019

    Istanbul Technical University

    An econometric analysis of the determinants of electricity prices in Turkish energy markets

    Question
    What actually moves the spot price of electricity in Turkey?
    Method
    A VAR model over 8,808 hourly observations, with Johansen cointegration, Granger causality tests and impulse response functions.
    Finding
    Natural gas, coal and Brent prices push the spot price up; hydropower production pulls it down.
    • VAR
    • Granger causality
    • Cointegration
    • Time series

07Signal received

Reference

I myself personally, as his direct manager over the last year, would hire Omer again, whether it be in my own business or elsewhere!

…an integral part of the engineering and products team.

Wojtek Bronikowski

Manager, Software Engineering · Songkick · Warner Music Group

08Contact

Let’s buildsomething that lasts.

I am looking for a long-term in-house role in data or AI engineering. If that sounds like your team, write to me.

Email
omerfarukballi@gmail.com
Phone
Available on request
LinkedIn
linkedin.com/in/omerfarukballis
GitHub
github.com/omerfarukballi
Résumé
Download PDF

New messageto omerfarukballi@gmail.com

Send opens your own mail app with this message ready to go.