Nada Sadek · Senior Software Engineer
My work focuses on billing, payments and account access.
I trace failures across retrieval, context construction and generation then define fixes, regression checks and reliability tooling.
Focus
Billing, payments and account access
Work
Outside-in assessment and internal investigation
Experience
10+ years in software engineering
AI Support Reliability
I compare the assistant's answers with the documented policy, help-centre guidance and product behaviour for that workflow.
I examine one workflow through the public support surface customers use. The report records the questions, answers, screenshots, source references, findings and severity.
Discuss an assessmentIncluded in the report
I define one narrow workflow such as a failed payment, subscription change or account-access issue.
I run realistic customer questions and compare each answer with the accessible source of truth.
You receive a report with the evidence, findings and recommended next steps.
Internal AI Support Reliability
I inspect retrieval results, context blocks, model inputs, outputs and traces to locate the first demonstrated failing stage.
From there, I define the remediation, regression checks and observability needed to verify the fix and catch similar failures.
This work can include
Retrieval evaluation · context construction review · evidence coverage · regression testing · review workflows · observability tooling
Work
Prototypes and technical projects in retrieval evaluation, evidence review and support tooling.
A retrieval evaluation tool for testing whether a help-center assistant finds the right source documents before generating an answer.
It compares keyword, embedding and hybrid retrieval strategies against a fake SaaS help-center dataset, with metrics for hit rate, recall and ranking quality.
A frontend QA prototype that turns screenshots, DOM snippets and accessibility findings into structured review items.
The project focuses on separating visual evidence from code evidence, so AI-generated findings stay inspectable instead of becoming vague bug reports.
A review queue for support messages that may affect billing, access or revenue.
It parses messy customer messages into structured risk signals so a human reviewer can quickly inspect the issue, correct the output and decide what needs attention.
About
I'm a senior software engineer based in Berlin with more than ten years of experience across SaaS, ad tech and gaming.
My background includes frontend architecture, backend integrations, support tooling, billing flows and complex product workflows.
Contact
Describe the current system, the workflow and what you need to understand or improve. I will reply with how I would approach it.
I also take selected senior software engineering contracts involving AI product systems, support tooling, billing and complex operational workflows.
Prefer email? nada@nadasadek.com