JiříPokorný

AI Engineer · Brno, CZ

I build AI applications in Python, from raw data to a deployed product.

Scroll
01Selected Projects
01 / 02

Master's Thesis

Authorship Verification System

End-to-end authorship verification system for English texts. Several hundred experiments across BoW/TF-IDF, graph-based, transformer (Longformer) and LLM approaches, with Optuna searching through thousands of preprocessing, feature and model pipelines. Best result F1 92.38%. The live web app (FastAPI + React) serves a lightweight logistic-regression model, small and cheap to run.

Authorship Verification System preview
PythontransformersLongformerOptunaFastAPIReactTypeScript
02 / 02

Bachelor's Thesis · SAP

ATACA Incident Classification

Binary classification of SAP customer incidents at sentence level, built on real production incident data. Systematic comparison of dataset balancing methods, vector representations and classifiers, traditional vs. modern approaches.

Pythonscikit-learnTF-IDFNLPSAP
02Tech Stack

Languages

PythonC++TypeScript

AI & NLP

scikit-learnNLTKspaCygensimtransformersOptuna

Data

pandasNumPySciPyPostgreSQL

Backend & Web

FastAPIDjangoReact

Cloud & DevOps

AzureDockerKubernetes

Tools & Practices

GitLinuxCI/CD
03Experience

Associate Developer

SAP · Brno, Czech Republic

Enterprise Apps

SEP 2024 - FEB 2026 · Full-time

Developed features for SAP S/4HANA applications across the backend (ABAP) and frontend (SAPUI5, Fiori). Debugged and resolved production issues in live customer systems.

ABAPSAPUI5Fiori

ML Engineering

JUN 2022 - SEP 2024

Built an NLP classifier that classifies SAP customer incidents, trained on real production incident data. This work became the foundation of my bachelor's thesis (see ATACA in Selected Projects). Alongside the thesis work, I handled other data and modelling tasks for the team.

PythonNLPscikit-learnText Classification
04Contact