Yigit Erturk
Open to remote product engineering work

I build systems that can show their work.

AI products, operational software, and data-heavy tools with explicit evidence, failure boundaries, and interfaces people can actually use.

Inspect selected work

19ClinEntra verification flows

27VoiceOps automated tests

14Open-source engineering skills

Reproduced15 Aug 2026

Selected work / 01—03

Products with receipts.

01 / Clinical intelligence

ClinEntra

An evidence-first workspace that detects field-level changes in public clinical-trial records and carries the source into human-reviewed operational decisions.

  • Tenant-scoped action, feasibility, and study-operations workflows
  • Deterministic verification for isolation, backups, delivery, monitoring, and secret hygiene
  • Bounded pilot—not clinical decision support or outcome prediction
Live controlled pilot ↗
ClinEntra intelligence brief showing evidence-linked clinical-trial changes

02 / Voice operations

VoiceOps Studio

A multilingual voice-operations prototype that turns customer conversations into structured, metered, consent-aware workflows across ten locales.

  • Fish Audio speech with explicit degraded and deterministic fallback modes
  • Signed Twilio flows, encrypted records, active-voice quotas, and provider readiness
  • 27 tests, type-check, and production build reproduced locally
Open live demo ↗ Source & case study ↗
VoiceOps Studio completed English appointment conversation

03 / Developer tooling

Engineering Skills

Fourteen evidence-first Codex skills for architecture mapping, production-readiness, tenant isolation, CI hardening, migrations, AI trust boundaries, and operational handoffs.

  • Dependency-free Python scanners with fixtures and claim boundaries
  • 19 unit tests reproduced; CI and CodeQL included
  • Open-source under MIT with an installer and contribution guide
Explore all 14 skills ↗
Engineering Skills social preview reading Evidence over adjectives

Open source / 04—06

Small tools for harder trust boundaries.

Three focused utilities extracted from the way I review and ship AI-enabled software. Each repository includes tests, CI, an explicit limitation section, and a reproducible local workflow.

04 / Repository evidence

ProofRepo

An evidence-first CLI that maps README, tests, CI, licensing, reproducibility, env hygiene, visual proof, and collaboration signals without pretending to certify the repository.

  • Markdown and JSON reports
  • Configurable CI exit policy
  • 3 tests; public CI passing
Inspect ProofRepo source and checks ↗ 
$ proofrepo .

score    98 / 100
pass     12
warn      1
fail      0

boundary evidence ≠ certification

05 / LLM reliability

LLM Output Contract

A provider-independent TypeScript library and CLI that extracts JSON from mixed model text, rejects unsafe object keys, bounds input and nesting, and validates the result against JSON Schema 2020-12.

  • Raw, fenced, and balanced extraction
  • Rejected values omitted from errors
  • 9 tests; public CI passing
Inspect the contract boundary ↗ 
contract.parse(modelText)

{
  ok: true,
  source: "fence",
  schema: "valid"
}

06 / Secret hygiene

SecretProof

A local-first Git secret scanner for worktrees, staged diffs, and bounded history. Findings expose a rule, location, severity, and one-way fingerprint—never the matched credential value.

  • Worktree, staged, and history modes
  • Text, Markdown, JSON, and SARIF
  • 8 tests; public CI passing
Inspect SecretProof's threat boundary ↗ 
$ secretproof . --mode staged

scanned    22 files
findings    0
high        0
medium      0

matched values are never printed

How I work

Calm interfaces. Explicit boundaries underneath.

  1. 01

    Model the failure states

    Authorization, provider outages, destructive actions, and unknown evidence are product states—not footnotes.

  2. 02

    Make claims reproducible

    Builds, tests, screenshots, contracts, and operational limits stay close to the code reviewers can inspect.

  3. 03

    Ship the whole path

    I work across interaction design, TypeScript and Python services, data models, CI, deployment, and handoff documentation.

Remote / worldwide

Have a complicated workflow that needs to become a real product?

I’m open to product engineering roles, focused collaborations, and evidence sprints. Scope first; pricing follows the actual problem.

Start a conversation on LinkedIn ↗ Inspect GitHub ↗