Senior MDM Architect

Job Title: Solutions Architect OR MDM Architect

Location: Hyderabad-- 100% Remote

Full Time

Job Description

We are seeking an offshore MDM Solution Architect to support the design and implementation of an enterprise Master Data Management program across customer and product data. This person will help define the overall solution architecture, guide technical delivery, and partner with offshore engineering resources to build a clean, trusted, and scalable data foundation.

This role is best suited for someone who can operate as both a strategic architect and a hands-on technical leader. The ideal candidate has experience designing MDM solutions, working with complex enterprise data, and providing direction to data engineers building pipelines, data models, match/merge logic, and golden records.

The Role:

You will design and build the customer master data model for an organisation resolving aconsumer-scale person universe and a large institutional estate across many source

systems, into a governed party model with reversible merges and attribute-levelsurvivorship. The model exists on paper and partly in SQL. It needs someone who can takeit from written to running, decide what is wrong with it, and own the matching logic that

determines whether the whole thing is trustworthy

Required Skills & Experience

- Experience as a Solution Architect, Data Architect, MDM Architect, or senior technical lead on enterprise data or MDM programs.

- Strong understanding of Master Data Management concepts, including match/merge, survivorship rules, golden records, cross-reference tables, hierarchy management, data stewardship, and governance workflows.

- Hands-on experience with Snowflake, SQL, and Python in a data engineering or enterprise data platform environment.

- Experience designing layered data architectures such as RAW / STAGED / CURATED or Bronze / Silver / Gold.

- Strong data modeling experience across relational, dimensional, and hierarchy-based models.

- Experience with data profiling, data quality rules, deduplication, reconciliation, and exception handling.

- Experience working with SAP as a source system, ideally involving customer, material, vendor, or product master data.

Nice to Have Skills & Experience

- Experience with commercial MDM tools such as Reltio, Informatica MDM, Semarchy, Profisee, Ataccama, or similar platforms.

- Experience integrating Salesforce data into an enterprise MDM or analytics platform.

- Exposure to graph modeling, entity resolution, AI-assisted matching, embeddings, vector search, or Snowflake Cortex AI.

What you would own

The party model

A single party root holding schools, districts, clubs, churches, colleges, coaches, and consumers with equal standing, where "customer" is a role rather than an entity. Subtypes, descriptors, crosswalk, lineage, and merge events. You will inherit an existing design and be expected to argue with it.

Probabilistic matching, and its quality

Blocking strategy, scoring model, thresholds, negative evidence, and vetoes tuned against an existing set of labelled links held as ground truth. There is a measured baseline for blocking recall and match precision, and improving it is the single highest-value technical outcome of the role.

Person resolution at consumer scale

The large majority of the person universe has no reference anchor, so consumer identity is self-merge on contact points under the constraint that a substantial share of source email addresses reach more than one account, some of them reaching thousands. This is the hardest problem in the programme and the reason the role exists.

Survivorship and rule governance

Per-attribute survivorship, not per record. Sources split by provenance where one source is partly a mirror of another. A rule engine where rules are rows rather than code, with firing telemetry and a regression gate against the labelled set.

Hierarchies as edges

Several trees coexisting over the same nodes legal district structure, operational coaching affiliation commercial rollup with a governed edge vocabulary that determines Senior MDM Architect coaching affiliation, commercial rollup with a governed edge vocabulary that determines which trees roll up revenue and which must not.

Required

These are the axes we will actually assess. Everything else is secondary.

Party / entity-relationship modelling as your primary skill

You can start from a blank page and reason about what is an entity, what is a relationship, where identity comes from, and what breaks under each choice. Ideally demonstrated outside a vendor context.

Probabilistic record linkage, hands-on

Blocking, candidate generation, scoring, thresholds, and the precision and recall trade-off stated separately per consuming use. You have tuned a matcher against labelled data and can describe what you changed and what it cost.

Person matching at consumer scale, under a privacy constraint

You understand that over-merging people is a disclosure rather than a reporting error, and that this asymmetry should set thresholds differently from organisation matching.

Attribute-level survivorship

Why record-level "source X wins" is wrong, how trust ranks fail when a source is partly a mirror of another, and why consent must never be survived at all.

Reversible merge design

Crosswalk-based merge where source records are never mutated, merge events are repayable in reverse, and consumers holding a merged-away identifier get a defined answer rather than an error.

Advanced SQL against large datasets

Window functions, recursive CTEs, closure tables, MERGE patterns, and performance reasoning at warehouse scale. You will write this code, not review it.

DBT, or equivalent transformation-as-code discipline

Tests, lineage, and version-controlled models. On a team these small, automated tests are the only reviewer available.

Reference-data reconciliation

Reconciling an internal estate against an authoritative external universe using external identifiers and understanding why a reference source must be de-duplicated before it is allowed to seed anything. Directness about what the data actually says Several findings in this programme came from querying the database and discovering that a documented step was never running. We need someone who measures rather than assumes, and who says so plainly when a number is wrong.

What you should know before applying

The build is real, partly running, and honestly documented. The master layer is designed and written but not yet fully deployed. There is an open

findings register that is actively maintained, a measured match baseline that is not yet good enough, and a golden-record seed that does not exist yet. Source systems disagree with each other in ways that are documented rather than resolved.

HONEST FRAMING

We are not presenting this as finished work. Someone who finds that interesting rather than alarming is the person we want and someone who

expects to arrive and configure a working product will be disappointed on the first day.

Mohd Safiuddin Khan

Lead Recruiter

HITEC City, Hyderabad, Telangana

[phone removed]

[email removed]

connect on LinkedIn

insightglobal.comRole & responsibilities

Preferred candidate profile