Hi, I’m Benedict

I’m a Computer Science master’s student at ETH Zürich, focusing on machine intelligence. My work spans AI security, robotics, and helping businesses put AI into practice through independent consulting. I previously graduated in the top 5% with a degree in Data Science and Artificial Intelligence from Saarland University.

  • M.Sc. @ ETH Zürich
  • AI and machine learning
  • Robotic Learning
  • Independent consultant
Benedict Böttger

Benedict Böttger Consulting · Since 2025

AI consulting & development

I help medium-sized businesses integrate AI into their work, from identifying useful applications to building automations and custom software. Based in Zürich, with consulting experience supporting German businesses.

  1. AI & workflow automation

    Connect AI tools to your existing processes and systems. From n8n workflows to custom integrations, I build automations around the way your team works.

  2. Data analysis & visualization

    Turn your data into insights you can act on. Explore patterns, understand performance, and communicate findings with clear visualizations.

  3. Custom software development

    Build the software that ties everything together, with a focus on Python, TypeScript, and applications powered by language models.

  4. AI competency training

    Together with my partners, I offer training to help teams work confidently with tools such as ChatGPT and Copilot, tailored to different experience levels.

Projects & publications

A selection of academic and hands-on work in robot learning, language models, and formal verification.

Course project · ETH Zürich · Spring 2026

Wall-E: a keyboard-typing robot

A robot that locates keys and types on a physical keyboard, built for ETH Zürich’s Robot Learning course. The project brought perception, motion planning, and control together on real hardware.

A fine-tuned YOLO model detects the keys, inverse kinematics translates their positions into arm movements, and PID control guides the key presses. Working with physical hardware meant debugging mechanical issues and connection failures alongside the software.

Robot learningYOLOInverse kinematicsPID controlPython

Team: Benedict Böttger, Niklas Britz, Peter Nutter, and Meghan Brandt.

Wall-E arm visualization and live keyboard camera view in the Rerun robotics interface

Bachelor’s thesis · Saarland University · October 2023

Synthesizing Verifiable Code for Large Specifications Using Few-Shot Learning

Can language models help synthesize digital circuits beyond the scale of classical synthesis tools? My bachelor’s thesis explores generating Verilog implementations from parameterized linear-time temporal logic (LTL) specifications, using solutions for smaller instances as few-shot examples.

For some benchmarks, the approach produces correct implementations at scales several orders of magnitude beyond classical tools alone, especially with human-written examples. The results point toward a hybrid workflow combining language models with existing synthesis and verification tools.

Author: Benedict Böttger
Supervisor: Prof. Bernd Finkbeiner, CISPA
Advisor: Frederik Schmitt
Thesis grade: 1.0 · Degree grade: 1.2

VerilogLTL synthesisFormal verificationFew-shot learningLarge language models
Specifications define the task, few-shot learning guides synthesis, and verifiable code is checked for correctness.

Academic interests

Machine Learning & Data Science

Python, Pandas, PyTorch, Matplotlib, Plotly, Jupyter Notebooks, Model Training

Robotics

Hands-on experience with robot learning, simulation, and real-world robotic systems through ETH's Robot Learning course; learning-based control and manipulation, simulation environments, and deployment on physical hardware.

Large Language Models

Exploitation of security vulnerabilities such as prompt injections (as well as countermeasures), Development of LLM-enabled apps

Let’s connect

Whether you’d like to discuss research, collaborate on a project, or find an AI solution for your business, I’m happy to hear from you.