Dhafer Bouthelja
I am Dhafer Bouthelja, a software quality lead with eighteen years in automotive and industrial electronics — from production floors and customer audits to ASPICE, ISO 26262 and ISO/SAE 21434 compliance on embedded software programmes.
qualitycrew.fr is my own project. I built it, I host it, and I run it.
18+ yearsin automotive and industrial electronics
5/6tools deliver their result without AI
Why this exists
Compliance work in safety-critical engineering is slow in a specific way: the standards are long, the evidence is scattered across tools, and the person reviewing a work product spends more time locating information than judging it. That gap is where I spend my days, and it is what these tools address.
Every tool here started as a problem I had actually met — a review that took three days when the reading took three hours, a regulatory change nobody noticed until an audit, a root-cause analysis that went in circles because the questions were asked in the wrong order.
AI never decides
This is the rule the whole platform is built on, and it is not a disclaimer.
Here, AI never rates and never decides what is kept. Five of the six tools produce their result without it; the only one that depends on it, QualityCrew, has its audit written by agents.
Decisions belong to a qualified human who remains accountable for them — that is what functional safety and quality management require, and no amount of model capability changes it.
In practice this means three things:
Every output is traceable. An agent that cites a clause points to the clause. An agent that raises a finding shows what it read to get there.
Every output is a draft. The tools produce material for a reviewer, never a signed conclusion.
Uncertainty is stated, not smoothed. An agent that lacks the evidence says so instead of producing a confident-sounding paragraph.
I would rather ship a tool that admits what it does not know than one that sounds authoritative and is occasionally wrong. In this field, the second kind is worse than no tool at all.
The tools
- QualityCrewmulti-agent compliance auditing against ASPICE and ISO 26262.
- RegWatchstandards and regulatory watch.
- CauseTrace8D root-cause analysis with 5 Whys, Ishikawa 5M and is / is-not questioning.
- SafetyScopeHARA risk analysis and ASIL determination (ISO 26262).
- ThreatScopeTARA threat analysis, through to cybersecurity goals (ISO/SAE 21434, UN R155).
- SentinelScaninformation-leak watch on public GitHub repositories, ranked by criticality.
The project
How it is built
A Python architecture, hosted on a server I administer, designed, developed, deployed and operated end to end by me. Agents are used where a model genuinely adds something, each with a defined scope. Everywhere else, the work is done by written, tested rules.
Scope
qualitycrew.fr is a personal project. It contains no company data, no client data and no confidential material of any kind. The standards are cited, never reproduced; the examples are my own.
Contact
linkedin.com/in/bouthelja-dhafer