What I led
Designed and built a natural-language-based SKU consolidation pipeline to normalize component naming and improve inventory signal quality across warehouses.
NLP SKU Resolution Engine
Warehousing and omni-channel logistics operations had fragmented SKU naming across systems, causing unreliable inventory views and inefficient routing decisions.
Stack
PythonNLP entity-disambiguation techniquesFuzzy matching/logic for product description normalizationLocal processing workflowSupporting ETL/data transformation and reporting layers
Highlights
- Built a pipeline to compare SKU descriptions and names across systems and identify equivalent component entities.
- Implemented disambiguation logic to map fragmented SKU variants to normalized canonical IDs.
- Enabled downstream inventory analysis to better represent true stock availability and cross-warehouse routing opportunities.
- Delivered a reusable solution pattern that was later sold/deployed across multiple client contexts.
- Needed to handle inconsistent naming conventions and noisy text fields across logistics data sources.
- Precision/recall tuning details and threshold configuration history are partially unknown due to project age.
- Implemented as a reusable capability/library rather than a directly user-operated production service.
Outcomes
- Improved inventory accuracy and representativeness across warehouse systems by consolidating SKU variants.
- Enabled better demand-routing decisions across warehouses through normalized inventory views.
- Achieved low-90s precision in disambiguation outcomes (recalled estimate).
- Solution pattern was sold to several clients.