Vienna, Austria
Hi, I'm Fedja Bogataj
I build full-stack applications and data-driven systems — from clean interfaces to ML pipelines.
I turn data and ideas into products people actually use.
I'm a full-stack developer and data science student based in Vienna, currently pursuing my Master's at TU Vienna. I enjoy working across the entire stack — from architecting backends to crafting clean interfaces.
My ML focus covers computer vision, RAG systems, and reinforcement learning. Outside of tech I'm involved with the Board of European Students of Technology, where I help organise international events across Europe.
Full Stack Developer
SRC d.o.o.
Built an end-to-end document management app for internal and cross-enterprise document sharing and signing. Designed the architecture, containerised with Docker, deployed on Azure, and integrated Active Directory SSO.
Software Engineer
SRC d.o.o.
Worked across the full stack (Vue.js + C#) on a Visitor Management System. Implemented camera-based sign-on, an interactive map widget, SMS/email notifications, and Google & Apple Wallet integration for digital visitor passes.
Application Tester
SRC d.o.o.
Performed various types of application testing, primarily using SoapUI.
MSc Data Science
Technical University of Vienna
My studies focus on two tracks — Machine Learning & Statistics and Big Data & High-Performance Computing — covering advanced regression and classification methods, Deep Learning for Visual Computing, Generative AI, statistical simulation, recommender systems, advanced database systems, GPU architectures, and parallel and data-intensive computing. I have parallelised compute-heavy workloads with OpenMP and written SLURM batch scripts for strong-scaling and scheduling experiments on HPC clusters. This combination equips me with both the theoretical foundations and the practical engineering skills to design, train, and deploy data-driven systems at scale.
BSc Computer Science
University of Ljubljana
My studies included topics such as Computer Architecture, Databases and programming in both C and Java, with a focus on Machine Learning and Computer Vision.
Teapot — AI Quality Control for Fiber-Optic Trench Documentation
ViennaUp Hackathon 2026
At the ViennaUp Hackathon 2026 I built Teapot, a computer-vision pipeline that screens hundreds of thousands of construction-site images for compliance, automatically verifying cable placement in the trench, trench depth, and on-site safety.
Industrial AI — Learning and Benchmarking Process Logic
Zero One Hack 2026
For the Industrial track of Zero One Hack 2026 I built transformer models that learn semiconductor fab process logic, performing next-step prediction, sequence completion, and anomaly detection across MOSFET, IGBT, and IC families. I trained and benchmarked them on a roughly 120-token vocabulary and thousands of synthetic sequences using Top-k accuracy, MRR, F1, and ROC-AUC, scheduling the training jobs via SLURM on the Leonardo EuroHPC cluster.
Full-member of BEST Vienna
Board of European Students of Technology Vienna
I joined BEST Vienna in March 2025, and became a Full member in December 2025. I was Logistics and Fund-raising responsible for beWANTED 2026 a career fair aimed at students at technical universities. I am also the logistics responsible for the BEST Summer Course 2026 in Vienna, a 10-day international event with 40+ participants from all over Europe, where I coordinate the logistics of the event and manage a team of volunteers to ensure everything runs smoothly.
My bachelor thesis explored deep reinforcement learning for financial market prediction, comparing feedforward, LSTM, and GRU agent architectures. Each agent learned to buy, hold, or sell directly from historical price data across multiple financial instruments over the 2019–2024 period. Results showed the feedforward agent performed best on most instruments, while the GRU agent had an edge on those with sharp, sudden price movements.
Notepilot is an AI-powered notetaking app that lets users capture notes and retrieve them by querying in natural language using Retrieval-Augmented Generation. An LLM-based summarisation layer distills longer notes into concise insights on demand. Built with Next.js and a vector database backend for fast semantic search.