Stanislav Guzej

Hi, my name is

Stanislav Guzej

I build ways to understand and trust AI systems.

I’m a final-year computer science student at Masaryk University. At Lexical Computing, I’m building an automated LLM testing suite as part of my bachelor’s thesis.

Experience

LLM Testing Suite @ Lexical Computing

Part-time · Current

  • Building an automated testing workflow for language-model systems as the practical part of my bachelor’s thesis.
  • Exploring repeatable ways to compare model behaviour, surface regressions, and make evaluation results useful to developers.
  • Turning experimental LLM workflows into systems that can be inspected, discussed, and improved.

Selected projects

Featured project · In progress

Automated LLM testing suite

A structured workflow for turning expectations about an LLM application into repeatable checks, helping developers notice behavioural change before it reaches users.

  • LLMs
  • Evaluation
  • Testing

Repository available after public clearance

Concept project · Draft

Scenario lab

A compact playground for comparing prompts and models against realistic user scenarios, with human review shown beside automated signals.

  • Python
  • Prompting
  • Analysis
Read the project outline
View all projects

Writing

Notes on LLM workflows, evaluation, and the practical details of building reliable systems. Notes on LLM workflows, evaluation, and the practical details of building reliable systems.

June 2026 · 4 min read

Evaluation that helps you decide

Why the most useful score is the one that changes the next engineering decision.

Draft
All writing

What’s next?

Let’s talk.

I’m always interested in thoughtful conversations about evaluation, LLM products, and difficult technical problems.

Say hello