I'm a PhD student in computer science. I work on large software systems and benchmarks, mostly in high-impact areas where better systems improve lives.

Get in touch ↗/Product Owner @ p36·Research & PhD @ TU Berlin
Available--:--:-- CET
Sebastian Koch
I work across AI safety, large-scale systems, and benchmarking, usually on whether software stays trustworthy as it grows.
AI Safety
Trustworthy LLM evals under GxP
Systems
Scalable systems that hold up
Benchmarks
Sensitive, reproducible performance
Impact
Systems that help where it counts
01

Work

Projects on GitHub →
gxpeval
A validation framework that lets regulated teams use AI accountably. It adds risk-tiered assurance, anytime-valid early stopping, qualified LLM judges, and a tamper-evident ALCOA+ audit trail, all on top of Inspect AI.
Open Source · Python
beforeyoudonate.org
Urging people to donate more effectively. Signatories include the co-founder of Lovable and Peter Singer.
Social Impact · Web
02

Publications

More on About →
Duet Instrumentation: An Agentic Approach to Improving Sensitivity in Cloud Service Benchmarking
Sebastian Koch, Nils Japke, David Bermbach
arXiv preprint · 2026PDF ↗Scholar ↗
Towards an Optimized Benchmarking Platform for CI/CD Pipelines
Nils Japke, Sebastian Koch, Hendrik Lukaszczyk, David Bermbach
IEEE IC2E · 2025PDF ↗Scholar ↗
Increasing Efficiency and Result Reliability of Continuous Benchmarking for FaaS Applications
Tim C. Rese, Nils Japke, Sebastian Koch, Tobias Pfandzelter, David Bermbach
IEEE IC2E · 2024PDF ↗Scholar ↗
03

Currently

Now →
Research Associate
TU BerlinTRANSFER — AI decision support for surgery
Product Owner
p36Cloud Service Qualification Platform (GxP)
Volunteer
Effektiv SpendenEffective giving