Fintech · Document Intelligence
Hypodossier: Mortgage Document AI
AI extraction and classification of mortgage documents, recognizing over 250 document types.
- Location
- Europe
- Duration
- 7 months
- Engagement
- Dedicated team (ML and backend)
Challenge & scope
Requirements
- 1Classify hundreds of document types from scanned mortgage files.
- 2Extract more than 500 data fields with validation.
- 3Keep all processing inside the EU.
- 4Integrate with the existing case management system.
“Automate the manual document administration behind mortgage applications.”
Key deliverables
- Document classification model
- Field extraction pipeline with validation rules
- Review UI for low-confidence fields
- EU-hosted processing infrastructure
- Monitoring and drift detection
Technical execution
Hybrid Extraction
Layout-aware models and LLM extraction are combined, with rules validating cross-field consistency.
Review Loop
Reviewer corrections become training data, which raised accuracy every month.
Data Residency
All processing runs in EU regions with strict retention policies.
Technical stack
ML
- PyTorch
- OpenCV
- Hugging Face
Backend
- Python
- Django
- Celery
Infrastructure
- AWS (EU)
- PostgreSQL
- S3
MLOps
- MLflow
- Grafana
Let's build your next platform together.
Get in touchFeatures & outcomes
Accuracy
95% field-level accuracy across 500+ fields.
Throughput
Processing time per file reduced from hours to minutes.
Coverage
Over 250 document types recognized.
Client benefits
- ML expertise embedded in the client team
- EU data residency handled end to end
- Measurable monthly accuracy gains
- Transparent model documentation
Strategic outcomes
- Faster mortgage decisions
- Lower processing cost per application
- A reusable document AI platform
- Competitive advantage in a regulated market
Project team
Roles on the Menna AI side of this engagement.
ML Lead
Owned model strategy and evaluation.
Backend Engineers
Built pipelines and integrations.
QA Engineer
Maintained validation datasets.
Start a conversation
Let’s solve a real problem together.
Whether you need software built, a network installed or a partner for a digital transformation program, your first conversation is with the people who will do the work.
We reply within one business day. Based in Addis Ababa, working with partners in Ethiopia and beyond.
- 2026
- Founded in Addis Ababa, Ethiopia
- 2
- Focus areas: software, networking & IT
- 4
- Core values: Service, Wisdom, Stewardship, Excellence
- 3
- Founding leadership team