Work 03RAG, QA and evaluation, production

Adalah

An AI-powered legal platform: retrieval-augmented generation over Saudi regulatory and policy documents.

Role
AI Engineer, part-time, at Adalah under VR House.
When
Mar 2026 to Jun 2026
Adalah on a laptop: the Arabic landing page
  • 3Measures: accuracy, citations, hallucination
  • 30+Defects triaged across the platform
  • 1:1Every failure became a test case

The role

I worked part-time as an AI engineer at Adalah, under VR House, in Al Khobar, evaluating the production legal assistant.

What I did

I evaluated the assistant and tested the platform around it.

  1. Scored the assistant's answer accuracy, legal-citation correctness and hallucination rate.
  2. Converted every failure into a reproducible test case that drove model-quality fixes.
  3. Contributed to retrieval quality and ran end-to-end QA across the RAG pipeline, classifying defects by system layer.
  4. Validated the API and database layers with Postman and PgAdmin, triaging 30+ defects across access control, input validation and file handling.

Stack

Field
RAG, QA and evaluation, production
Stack
  • RAG
  • Saudi Regulations
  • QA & Evaluation
  • Production
Links