EMIR HÜSEYİN İNCİ / SYSTEMS & DATA INFRASTRUCTURE

Emir Hüseyin İnci

Systems & Data Infrastructure Engineer

I design and build deterministic execution kernels, Arrow-native data tools, and low-latency runtimes where correctness is proven by construction.

A / CORE ARTIFACTS

Three systems. Each with a reproducible boundary.

source / method / result

01 / STABLE / 0.6.1

Deterministic, fixed-point decision kernel.

catalog + policy + request → decision + audit bundle

$ cargo add calybris-core@0.6.1

02 / ALPHA / 0.7.2

Arrow-native contract and verification engine.

Arrow batches + contract + limits → report + evidence

$ pip install proofframe==0.7.2

03 / ACTIVE ALPHA / 0.1

Hierarchical delta debugging and AST reduction.

failing project + oracle → verified minimal reproducer

$ cargo install --git https://github.com/emirhuseynrmx/reprocut reprocut-cli --locked

B / ENGINEERING NOTES

question / method / finding

7,358,400 archived forecasts scored against a guess that needs no model, across thirty Turkish cities and two years.

Eight measurements closing a search for an edge in FX bars, including the three times a measurement caught me believing something that was not there.

A released page reported the previous version. The served bytes were correct and the cache-busting stamp was working exactly as written.

The bundled churn dataset records no treatment assignment, so uplift computed on it describes the generator rather than the customer.

Four decisions about the critic stage of an agent pipeline, and what each one costs. The weak one is marked as weak.

C / STATUS

16 Sep 2026

calybris-core  0.6.1  ~115 ns / decision
proofframe     0.7.2  Airflow + Dagster, browser WASM
reprocut       0.1    3 / 3 final verification

Claims name the code path, fixture, and boundary. No end-to-end latency is implied.

Correctness is not a claim at the end of the page.
It is an invariant carried through every decision, byte, and replay.