Remote — Germany & South Africa

Tina Koziol

I'm a data science and AI professional with 10+ years of experience spanning the evolution of the field, from debugging Python on Stack Overflow to harnessing cutting-edge coding agents like Claude and Codex. More recently, my focus has been on enterprise AI adoption, helping organisations go from pilot to production with governance frameworks, safety guardrails, and pipelines that don't fall apart at scale. My PhD simulated 240,000 interacting agents on HPC clusters, and that systems thinking still drives how I design agentic AI, verification loops, and trustworthy automation. I help organisations turn AI from promising pilots into governed, operational capability. The shift AI is bringing is nothing short of transformative, and I want to be where it's built, not where it's watched.

Open to roles PhD Economics,UCT EU AI Act / NIST RMF DE / EN / FR
Get in touch Work
Tina Koziol
Tina Koziol
Enterprise AI · Data Science · Governance
Experience

Complex systems, messy data, real deadlines.

Complex systems, real stakes, and getting AI to work inside organisations — not just in demos.

2021,2025
Hamburg
4 years
Senior Data Scientist
Wintershall Dea GmbH

Ran AI/ML projects from idea through production, including an LLM document pipeline (Claude, OpenAI) that hit 99% extraction accuracy on messy, real-world files.

Led cross-functional teams and mentored junior data scientists through production delivery.

Built the MLOps layer — CI/CD with GitHub Actions, monitoring, model governance — so nothing stayed stuck in a notebook.

Technical implementation of an LSTM-based architecture on well sensor data to support detection of well integrity issues.

Owned data curation pipelines and quality frameworks — because models are only as good as what goes in.

Led EU AI Act and NIST AI RMF alignment across 120+ staff. Not a checkbox exercise — actual workflow redesign, risk classification, and safety guardrails.

2020,2021
Remote / SA
1 year
Computational Researcher
GEMS — Univ. of Minnesota & Univ. of Stellenbosch

Large socio-economic datasets, cross-institutional research, policy-facing outputs.

Python, SQL, Postgres, GIS. Worked across time zones and disciplines to support food security modelling.

2016,2020
Cape Town
4 years
PhD Researcher & Lecturer
University of Cape Town — School of Economics (AIFMRM)

Built agent-based models with 240,000 agents on HPC clusters to study how financial crises spread.

Published on systemic risk and contagion — peer-reviewed, cited, still relevant.

Taught econometrics and programming. Supervised student projects.

2014,2016
Johannesburg
2 years
Consultant
Econometrix Ltd

Market studies for South African energy and finance. Budget management, profitability analysis, client delivery.

Career Trajectory
2014 Consulting 2016 PhD Research 2020 Comp. Research 2021 Sr. Data Scientist
10+
years in data
PhD
computational econ
2
continents
Skills

The tools, and how I use them.

Where I've spent my time
Python / Data10+ yrs LLMs / Agentic AI4 yrs Enterprise AI adoption4 yrs AI governance / safety3 yrs MLOps / Cloud4 yrs Research / agent-based models6 yrs Consulting / strategy2 yrs

Programming

PythonDaily driver for a decade. Data, web, automation, ML.
REconometrics, research, statistical modelling.
SQL / PostgresComplex queries, migrations, production schemas.
Django / WebTwo live products built end-to-end with Django.

ML & AI

Claude / Codex / OpenAIProduction pipelines, not just chat prompts.
RAG / NLP / Agentic AIDocument intelligence, extraction, verification loops.
Anomaly DetectionLSTM on real sensor data. Deployed, not just prototyped.
Agent-Based Models240k agents on HPC. Published in JEBO.

Cloud & MLOps

Azure AI / MLModel deployment, monitoring, governance.
Docker / CI/CDContainerised pipelines, reproducible builds.
GitHub Actions24 automated workflows running in production.
HPC ClustersLarge-scale simulations, parallel computing.

Governance & Leadership

Project ownershipEnd-to-end delivery, no hand-holding needed.
Cross-functional teamsEngineers, domain experts, leadership — kept aligned.
EU AI Act / NIST AI RMFLed compliance across 120+ staff. Not a checkbox exercise.
Data curation / guardrailsQuality in, quality out. Built the frameworks.
⚙

Builder

Dashboards, workflows, websites, internal systems,I make things and keep improving them.

⚖

Governance-minded

EU AI Act, NIST AI RMF, safety guardrails, data curation. I think about what goes into a system before asking what comes out.

⇄

Systems thinker

My economics PhD trained me to see interactions, incentives, and emergent behaviour. That still shapes everything.

▶

Driver

I take ownership, make decisions, and move work forward. I don't wait to be told what to do next.

Selected work

Proof beats promises.

Enterprise AI

Document Intelligence Pipeline

Extracts structured information from messy corporate documents. Not a demo — ran in production, used by people who didn't know it was AI.

AzureOpenAIClaudeLangChainPython
Research

COVID-19 Network Model

Agent-based model with 240,000 agents simulating how social learning and contagion interact in complex networks. Published in JEBO.

PythonRHPC240k agents
Safety AI

AI for Well Integrity Monitoring

LSTM architecture on well sensor data to support detection of well integrity issues. Built on real operational data, presented at ADIPEC/SPE.

LSTMWell IntegrityMLOpsAzure
Tooling

AI Model Atlas

Comparison site for AI model pricing and capabilities. Automated weekly refresh via GitHub Actions, source-linked.

DjangoGitHub ActionsPythonData curation
Venture

capetowndata.com

A multilingual information site for Cape Town — safety data, neighbourhood guides, local knowledge. Thousands of monthly readers.

DjangoPythonCloudflareStripeMultilingual
Venture

bubble-tracker.com

Tracking and visualising economic and market bubble indicators — turning scattered signals into something you can watch over time.

DjangoPythonData vizEconomics
Publications

Peer-reviewed work.

2024
EAGE Digital — Conference
MLOps Design Principles for E&P
How to set up AI infrastructure and governance frameworks in a large energy company so they don't collapse six months after launch.

Education

PhD in Economics
University of Cape Town,2016–2020
Computational economics, agent-based models, HPC simulations. Thesis on systemic risk and contagion.
Diplom-Kauffrau
Friedrich-Schiller-Universität Jena,2013
Equivalent to a master's in business administration.

Languages

GermanNative
EnglishFluent
FrenchIntermediate

Get in touch

Email me LinkedIn