Regulatory technology
Quanta Risk
Screening that stops guessing
- Sector
- Regulatory technology
- Location
- Dubai, UAE
- Year
- 2025
- Kickoff to production
- 15 weeks
Sanctions and AML screening rebuilt around entity resolution, so analysts review real matches instead of spelling variants.
Legacy screening flagged nineteen names for every genuine hit. Analysts cleared alerts under time pressure, and when the regulator asked why a customer was cleared in March, nobody could reproduce the decision.
- 01
Resolve the entity, then match
Transliteration, corporate structures and known aliases resolve into a single entity before screening, which removes the bulk of duplicate alerts outright.
- 02
Risk tiers, not a single queue
Alerts are scored and routed. Low-risk patterns auto-clear with evidence attached; genuine ambiguity goes to a senior analyst with context already assembled.
- 03
A case file that replays
Every decision stores the data, list version and model version used, so any past outcome can be reconstructed exactly.
False positives fell sharply and the team passed its next regulatory review with the case files as the primary evidence.
What changed
- Screening matches
- Entity-resolved
- Case files
- Replayable
- On real hits
- Analyst time
Stack
- Python
- Postgres
- Elasticsearch
- Next.js
- Airflow
Disciplines applied
- AI solutions
- Data & analytics
- Financial technology
