I test systems that don't give the same answer twice.
Senior SDET / QA Engineer, 18+ years. Most of my work now is on software that is
probabilistic by design — LLM agents, AI pipelines, quantum circuits — where the
usual tools of the trade quietly stop working. assertEquals has nothing to say
about a system whose correct output is a distribution, and a threshold someone
picked by feel is not a test. So I build the layer that does: deterministic
protocol-level assertions where behaviour is pinned, calibrated statistical
validation everywhere else.
Statistical assertions for quantum circuit tests. Chi-square goodness of fit, exact binomial and equivalence tests, power analysis, and suite-wide false discovery rate control so a large statistical suite doesn't go flaky on its own alpha. Zero runtime dependencies — the incomplete gamma and beta functions, the central and non-central chi-square and the Wilson intervals are implemented from scratch and verified against SciPy. The library tests its own calibration.
Python · hypothesis testing · pytest plugin · Qiskit / Cirq compatible
🛰️ FermiParadox
Civilization simulator exploring why a universe this large looks this empty.
Rust · React · Three.js
Relativistic travel simulator — time dilation and length contraction, rendered.
React · TypeScript · Three.js
Testing Playwright · pytest · Selenium · Cypress · Postman · TestRail
Languages Python · TypeScript / JavaScript · SQL
AI evaluation LLM-as-a-judge calibration · agent trajectory testing · Azure OpenAI · semantic embeddings
Infrastructure GitLab CI · Jenkins · AWS (Lambda, S3, Athena) · Git · Linux
I publish on testing non-deterministic systems — what breaks when you point traditional QA at an LLM, and what to do instead. LinkedIn →
English, Portuguese and Hebrew
