Hello! 👋

I'm Sebastian Frey

Applied ML Engineer / Data Scientist

About Me

I work at the intersection of machine learning, LLMs, and real-world business problems, with a strong focus on applied modeling, forecasting, and production-ready systems. My background combines deep technical work with business-critical use cases. As a Data Scientist, I design and implement advanced ML models end to end — from raw, large-scale data pipelines to evaluation, validation, and deployment — while keeping a clear view on how models are actually used for decision-making.

Python (Advanced) PyTorch & JAX Graph Neural Networks (GNNs) NVIDIA H100/A100/T4 Orchestration CUDA Optimization Spark & Polars GCP SQL

Projects

01

LLM-Augmented GNN Forecasting (Master's Thesis)

Developed a novel spatiotemporal framework using LLM-derived semantic graphs to predict Cost-Per-Click (CPC) auctions.Proved that explicit relational structures improve medium-to-long-term robustness over foundation models.

PyTorchGNNsLLMsSentence-BERTDuckDB
02

4youcast: Personalized Generative Audio

Architected a 'zero-caching' real-time pipeline for personalized news-to-podcast generation. Orchestrated complex LLM workflows and Google Cloud TTS synthesis for high-fidelity unique audio output.

PythonPostgreSQLGCPGitHub ActionsReact
03

Nationwide Real-Estate Valuation API

Designed and deployed a model covering 1.8M+ records with 300+ features, providing base price estimations for 80% of the Austrian market via a real-time production API.

Time SeriesR-ShinyMixed ModelsData Quality Monitoring

Experience

Co-Founder & Lead Engineer

4youcast

Jan 2025 - Present
  • Building a personalized news-to-podcast platform using AI-powered assistant pipelines.
  • Implemented robust API orchestration and rate-limit handling for LLM-based workflows.
  • Developed a zero-caching architecture to handle unique, user-specific audio generation.

Data Scientist / Product Owner

IMMOunited

Nov 2023 - Present
  • Engineered nationwide real-estate price prediction tools utilizing mixed models and time series.
  • Managed 1.8M records and 400 features to power internal and external client pricing tools.
  • Served as the technical interface between business stakeholders and engineering for production ML systems.

Risk Consultant (Financial Services)

KPMG Austria

Jul 2021 - Jun 2023
  • Validated risk models for national and commercial banks with assets up to €120bn.
  • Developed the first Sanction Risk Assessment Tool for the Austrian market prior to the 2022 crisis.
  • Led 2-4 person workstreams delivering analytics-driven reporting and stress testing.

Education

Master in Business Analytics

Nova School of Business and Economics

Aug 2024 - Jan 2026
  • Nova SBE Merit Scholarship recipient.
  • Master's Thesis: LLM-augmented time series forecasting at scale using GNNs.

BSc Business & Economics (Data Science Major)

Vienna University of Business and Economics

2020 - 2023
  • Merit Scholarship recipient.
  • Thesis focused on predicting residential rent through data science models.