DUPLICATE RECORDS
Same entity, different record
Detect exact and near-duplicate records created by repeated entry, revisions, parallel systems or inconsistent formatting.
SPECIALIST AREA
Entity Resolution & Matching.
I identify records that refer to the same real-world customer, company, product, property or other entity (even when the records are not identical) and reconcile them into reliable entity-level data.
Python
Record Linkage
Fuzzy Matching
Random Forest
Graph Analytics
Correlation Clustering
Deduplication
THE PROBLEM
Different records do not always mean different entities.
The same customer, company, product, property or asset can appear several times across a dataset or across different systems. Small changes in names, attributes, formatting or source information can make those records difficult to connect. I build entity-resolution workflows that identify these relationships without relying only on exact matches.
WHAT I CAN RESOLVE
From obvious duplicates to records that only partially agree.
CROSS-SOURCE MATCHING
Link records across systems
Connect records from separate files, databases, suppliers or platforms when no reliable shared identifier is available.
NOISY ATTRIBUTES
Match through imperfect information
Use similarity across names, prices, addresses, categories and other attributes when individual fields disagree or evolve.
CONFLICTING RELATIONSHIPS
Reconcile inconsistent matches
Resolve cases where pairwise evidence creates contradictory or incomplete relationships before records are collapsed into entities.
WHY ENTITY RESOLUTION
Pairwise matches still need to become coherent entities.
PAIRWISE EVIDENCE → ENTITY-LEVEL DECISION
Matching two records is only part of the problem.
Similarity rules or machine-learning models can produce relationships that do not fit together cleanly. One record may match a second, the second may match a third, while the first and third appear different. I use graph-based reconciliation to turn pairwise evidence into internally consistent groups representing the underlying entities.
A
B
C
D
likely same entity
likely different
RAW RECORDS
4
→
RESOLVED ENTITIES
2
HOW I WORK
From candidate records to traceable, entity-level data.
01
DEFINE THE CANDIDATE SPACE
Reduce impossible comparisons early.
Identify which records can plausibly refer to the same entity and define blocking or candidate-generation rules that keep the comparison space useful and computationally manageable.
02
BUILD PAIRWISE EVIDENCE
Measure similarity instead of requiring exact equality.
Compare relevant fields and score candidate pairs using rules, fuzzy similarity measures or supervised classification when the problem requires it.
03
RESOLVE RELATIONSHIPS
Turn pairwise predictions into coherent groups.
Inspect graph structure, detect contradictory relationships and apply an appropriate reconciliation rule so that the final entity assignments remain internally consistent.
04
DELIVER FINAL ENTITIES
Keep the resolution traceable.
Deliver a deduplicated or linked dataset together with entity IDs, source-to-entity mappings and reusable matching logic where appropriate.
TYPICAL PROJECTS
When several records may describe the same real-world thing.
01
CUSTOMER & CRM DEDUPLICATION
Find repeated customers or organisations hidden behind inconsistent records.
Identify duplicate customer, account or company records created through spelling differences, repeated entry, incomplete fields or disconnected systems.
02
PRODUCT & CATALOG MATCHING
Match equivalent products across suppliers, feeds or catalogues.
Link records that describe the same product despite differences in naming, category structures, specifications or source-specific identifiers.
03
MARKETPLACE & LISTING DEDUPLICATION
Separate repeated advertisements from genuinely different items.
Resolve listings created by reposting, parallel sellers, revisions or multiple representations of the same underlying property, product or offer.
04
MASTER DATA RECONCILIATION
Build one consistent entity layer across fragmented databases.
Link companies, vendors, assets or other recurring entities across separate sources and preserve the mapping between source records and the final resolved entity.
HAVE A MATCHING PROBLEM?
Let’s figure out what belongs together.
Tell me what records need to be matched, deduplicated or reconciled. I can help turn fragmented data into a traceable entity-resolution workflow.