~/ben-tischberger$

Ben Tischberger

Computer Science master's student at the University of Tübingen, working part time as a software developer in Paul Horn's Digital Solutions department, focusing on machine interfaces and production data transfers.

I am interested in LLM agents and their evaluation, machine learning and reinforcement learning.

CVThesisGitHub

Selected Projects

BoardBench

Bachelor's thesis, 2026

Evaluation of coding agents that generate Python game environments from board game rulebooks.

  • Workflow that extracts cited rule facts from a rulebook, resolves ambiguities and generates an implementation in an isolated workspace.
  • Separate evaluation layers: technical checks, runtime robustness, interface checks, cited rule scenarios and three independent blind LLM reviews.
  • Evaluated six games from German publisher rulebooks, comparing original generations with selected fresh generations given additional rule explanations.

StudyPlanner

StudyOS course project, 2025–2026

As part of Prof. Peter Gehler’s StudyOS course at the University of Tübingen, my team and I developed StudyPlanner during the 2025–2026 semester. The planner helps students find courses, plan semesters and track degree progress.

  • Searchable course catalog with schedules, semester plans, a weekly calendar and calendar export.
  • Import of the Transcript of Records PDF directly in the browser to track degree progress against the examination regulations.
  • AI integrations (OpenAPI and MCP) limited to reading the public course catalog, without access to account data.

Lilith

Personal prototype, 2026

Experimental chat and agent application for trying out memory and tool use.

  • Local alpha with tasks, memories and data retention rules.
  • Each tool call with external effects requires user approval bound to its exact arguments.
  • Web reading and browser tasks run in isolated Docker jobs with restricted network access.

Status

On hold during the semester.

Flying Watermen

Team project in Godot, 2023–2024

A game where you play a water slime trying to save a forest from fire creatures.

  • Melee, ranged and exploding enemies spawn across the map, and a boss waits at the source of the fire.
  • On the way there the slime finds old combat techniques and gets stronger.
  • I contributed gameplay components: projectiles, a wind tunnel trap and the burning wells.

Earlier projects

Education

10/2026 – present

M.Sc. Computer Science

University of Tübingen

10/2022 – 09/2026

B.Sc. Computer Science

University of Tübingen

2013 – 2021

Abitur

Gottlieb-Daimler-Gymnasium, Stuttgart

  • Peer mediator.

Professional Experience

03/2025 – present

Part Time Software Developer, Machine Interfaces

Paul Horn GmbH, Digital Solutions, Tübingen

  • FileAgent: Python service for machine backups and measurement data transfers, deployed at production sites in Germany, the US and the Czech Republic.
  • DncAppV2: OPC UA service supporting four CNC control types, running on all relevant US machines, with Germany following.
  • Drawing OCR: Local OCR service for technical drawings, built after benchmarking OCR engines and vision models.
  • Machine Upload: Parallel file distribution to up to 20 machines, with scheduling, conflict handling and FileAgent integration.

2023 – 2025

Computer Science Tutor

University of Tübingen

  • Supported programming exercises and reviewed assignments over three semesters.

2021 – 2022

Voluntary Social Year

District Town Hall, Stuttgart-Bad Cannstatt

  • Supported the Youth Council and helped organize local events.

2019 – 2023

Hospitality and Event Catering

Aramark, Stuttgart Spring Festival and Neckarmüller Tübingen

  • Hospitality and event catering roles alongside school and university.

2019 – 2020

Sports Coaching

Volleyball and table tennis

  • Volleyball school coach and table tennis trainer (D level).

Skills

Programming languages
Python, Java
Tools and systems
OPC UA, Windows services, REST APIs, Docker
Areas of interest
LLM agents, agent evaluation, machine learning, reinforcement learning, software engineering
Languages
German (native), English (C1, TOEFL iBT 2026)

Contact