Theme-aware portfolio: GitHub automatically serves the light or dark visual set based on your appearance setting.
I work on the proof layer between failing systems and safe engineering decisions.
My focus is simple: find the first real signal, explain what it proves, protect the review boundary, and attach the command or artifact that verifies the decision.
signal → failure vector → safety decision → proof command → reviewer-ready report
| Surface | What I look for |
|---|---|
| Python test behavior | correctness, hidden-test safety, regression signals, focused proof |
| AI-generated code | task logic, rubric alignment, hallucinated behavior, unsafe assumptions |
| CI and repository state | first failing line, failed step, owner files, exact next command |
| Data artifacts | Pandas tables, structured logs, JSON/Markdown evidence reports |
failure_vector:
check=<workflow/check>
command=<failed command>
first_failing_line=<exact signal>
failure_class=<formatter|lint|type|test|dependency|security|unknown>
safe_fix_candidate=<yes|no>
review_decision:
diagnosis=<what failed and why>
proof=<focused command>
blocked_actions=<what must not be changed>
next_human_action=<exact command or review step>
DevS69 SDETKit is my main public engineering project.
It is a local-first reliability platform direction for Python SDET workflows, CI diagnosis, failure evidence, proof artifacts, safety gates, trajectory memory, and PR-quality reporting.
Repository: https://github.com/sherif69-sa/DevS69-sdetkit
| Component | Product role |
|---|---|
| FailureVectorEngine | extracts the first real failure and turns it into structured evidence |
| SafetyGate | decides safe-fix eligibility and keeps unknown/risky work review-first |
| TrajectoryStore / RepoMemory | records action → response → diagnosis → proof → outcome |
| ReplayableBenchmarkHarness | evaluates no-op, oracle, and unsafe repair scenarios |
| ProtectedVerifier | independently checks patch scope, proof validity, and anti-cheat boundaries |
| PRReporter / PatchScorer | renders exact failure, safety decision, proof, next action, and repair score |
LeetCode is part of my programming discipline. I use it to strengthen algorithmic judgment, hidden-test safety, runtime awareness, memory tradeoffs, and debugging speed.
Profile: https://leetcode.com/u/sherif69/
| Track | Purpose |
|---|---|
| Python / Python3 | main algorithms and data-structure practice |
| Pandas | dataframe, table, log, and reporting practice |
| TypeScript | support track for typed APIs, closures, classes, and JS/TS-specific problems |
role_fit=Python SDET / AI Evaluation / Coding Benchmark QA
public_project=DevS69 SDETKit
working_method=evidence-first validation
practice_system=Python + Pandas + TypeScript
professional_signal=proof before claims
Python · pytest · Pandas · Docker · Git · GitHub Actions
CLI workflows · JSON evidence · Markdown reports · CI-style validation
AI evaluation workflows · coding benchmark review · repository inspection
- Python SDET
- AI Evaluation Engineer
- Coding Benchmark QA Engineer
- Test Automation Engineer
- CI Reliability / Quality Engineering
- LLM Workflow Evaluation
- Repository Review and Validation
- LinkedIn: https://www.linkedin.com/in/devs69/
- GitHub: https://github.com/sherif69-sa
- LeetCode: https://leetcode.com/u/sherif69/
- Email: sherif.atef6300@gmail.com