Data architect

What does a data architect do?

A data architect designs how data is stored, moved and used across an organisation.

You decide the structure of databases, data warehouses and data flows, and you set the standards that other data teams follow.

In Singapore, the role sits in the data or technology function of banks, government agencies, telcos, logistics firms and technology companies. It usually reports to a head of data, a chief data officer or a head of engineering.

You work with data engineers, analysts, data scientists, security teams and business owners. You explain designs to people who are not technical, and you check that designs follow privacy and governance rules.

What the work involves

• Map how data flows between systems and agree the target design with business and technology teams.
• Build and review data models for warehouses, lakes and operational databases.
• Write standards for naming, quality, access and security, and check that project teams follow them.
• Advise engineers on tool choices, such as a cloud warehouse, a streaming platform or a processing engine.
• Review how AI and machine learning models get deployed, and check that they meet governance and ethics rules.
• Present designs and trade-offs to stakeholders and update them as business needs change.

Pay in Singapore

Back to top

Skills a data architect needs

Back to top

Data architect job description

We are looking for a data architect to design and govern our enterprise data platform.

You will define how data is modelled, stored and shared across business units, and guide engineering teams as they build to that design. You will work closely with data, security and compliance teams. The role suits someone who can link business needs to technical choices.

Responsibilities

• Define the target data architecture and the way data should flow across systems and platforms.
• Design logical and physical data models for data warehouses, data lakes and operational systems.
• Set standards for data quality, metadata, lineage, access and retention.
• Choose and review platforms such as cloud data warehouses, batch processing engines and streaming tools.
• Align data designs with enterprise architecture and information security standards.
• Work with business owners to turn reporting, analytics and AI needs into data requirements.
• Check that designs follow data privacy policies, ethics guidelines and governance frameworks.
• Guide data engineers during build, and review designs and pipelines before release.
• Advise on how machine learning models move into production and how their data is managed.
• Explain designs, costs and trade-offs to senior stakeholders, and mentor junior team members.

Requirements

• A degree in computer science, information systems, data science or a related field is typical.
• Often 6 to 10 years of experience in data engineering, database administration or analytics before moving into architecture.
• Strong SQL and data modelling skills, including dimensional and normalised designs.
• Hands-on experience with a cloud data warehouse such as Snowflake or BigQuery, and with Spark, Databricks or Kafka.
• Working knowledge of data governance, privacy rules and security practice, and clear communication with non-technical staff.

Back to top

Data architect career path

Most data architects start in database, analytics or data engineering work and move into architecture after several years of building systems.

From there, you can lead architecture teams or move into data leadership.

1. Database administrator or data analyst

   Entry

   You learn SQL, database operations and how business teams use data. This builds the base for design work.

2. Data engineer

   Mid

   You build pipelines and warehouse tables, and learn how design choices affect cost, speed and quality.

3. Data architect

   Senior · SkillsFuture title

   You design the data architecture, set standards and guide teams on models, platforms and governance.

4. Lead or principal data architect

   Lead

   You own architecture across several domains, lead other architects and advise senior management on data strategy.

5. Chief data officer

   Head

   You set data strategy for the whole organisation and are accountable for governance, data value and team capability.

Steps marked "SkillsFuture title" follow the career pathway in the SkillsFuture Skills Framework. The others are titles commonly used in Singapore.

Related jobs and their pay

Back to top

Data architect interview questions

We wrote these questions from the skills this job needs. They are not an official question bank.

1. When would you choose a dimensional model over a normalised model, and how do you decide?

   What a good answer covers

   • Dimensional models suit reporting and analytics with simple joins and fast reads.
   • Normalised models suit transactional systems that need consistency and fewer update errors.
   • Mentions facts, dimensions, grain and slowly changing dimensions.
   • Considers the query patterns, the data volume and who uses the data.

2. Walk us through how you would design a data platform for a company moving from on-premise databases to the cloud.

   What a good answer covers

   • Starts with current systems, data sources, users and business goals.
   • Defines layers such as raw, cleaned and curated data.
   • Compares platform options and explains the trade-offs on cost, skills and lock-in.
   • Plans migration in phases with checks on data quality and security.
   • Names how access, lineage and ownership are managed.

3. How do Snowflake and BigQuery differ in how they store data and charge for compute, and how does that affect your design?

   What a good answer covers

   • Explains separation of storage and compute in both, and how each scales.
   • Notes the different billing models and how query patterns drive cost.
   • Covers partitioning, clustering and warehouse sizing choices.
   • Links design choices to cost control and performance.

4. A SQL query on a large fact table has become slow. How do you find the cause and fix it?

   What a good answer covers

   • Reads the query plan to find scans, joins and skew.
   • Checks filters, join keys and unneeded columns.
   • Considers partitioning, clustering, pre-aggregation or a model change.
   • Tests the change on realistic data and measures the result.

5. When would you use Spark on Databricks instead of SQL in the warehouse?

   What a good answer covers

   • Spark suits large-scale transformations, unstructured data and machine learning workloads.
   • Warehouse SQL is simpler and cheaper for structured, analytical work.
   • Mentions Delta tables, cluster cost and team skills.
   • Avoids adding a tool when the existing platform is enough.

6. Your team wants near real-time data from an order system. How would you design the flow using Kafka?

   What a good answer covers

   • Defines topics, keys and partitions based on how the data is read.
   • Covers schema management and how changes are handled.
   • Explains delivery guarantees, duplicates and replay.
   • Decides which data needs streaming and which can stay in batch.
   • Plans monitoring for lag and failures.

7. How would you set up data lineage, ownership and a data catalogue so teams can trust and find data?

   What a good answer covers

   • Assigns an owner and steward to each key data domain.
   • Captures metadata and lineage automatically where possible.
   • Defines quality rules and who is told when they fail.
   • Starts with the most important data instead of everything at once.

8. How would you use an AI coding assistant such as Claude Code in your architecture work, and what limits would you set?

   What a good answer covers

   • Uses it to draft table definitions, SQL and documentation, then reviews every output.
   • Tests generated code against sample data before it reaches production.
   • Keeps personal and confidential data out of prompts.
   • Keeps human accountability for design decisions.

1. Two business units define 'active customer' differently, and their reports do not match. What do you do?

   What a good answer covers

   • Meets both teams to understand each definition and its purpose.
   • Proposes one agreed definition, or two named metrics with clear labels.
   • Records the definition in a catalogue and in the data model.
   • Gets a senior owner to approve it so the dispute does not return.

2. A data scientist wants to push a model into production next week, but the training data has no clear owner. How do you respond?

   What a good answer covers

   • Asks where the data came from and whether its use is allowed.
   • Checks privacy, ethics and governance rules before deployment.
   • Finds or assigns an owner, and records lineage.
   • Offers a fast route, such as a limited release, instead of a flat refusal.

3. Warehouse costs have doubled in three months. How do you find out why and bring them down?

   What a good answer covers

   • Reviews usage by team, query and schedule.
   • Finds duplicate pipelines, full reloads and unfiltered queries.
   • Adjusts warehouse sizes, storage tiers and refresh frequency.
   • Sets budgets, alerts and usage rules so the problem does not return.

1. Tell us about a design decision that engineers disagreed with. How did you handle it?

   What a good answer covers

   • Describes the decision and the reasons for each side.
   • Shows that you listened and tested the concerns with evidence.
   • Explains the outcome and what you changed or kept.
   • Keeps the relationship with the team intact.

2. Describe a time you explained a technical data design to senior non-technical stakeholders.

   What a good answer covers

   • Starts with the business problem, not the tools.
   • Uses diagrams and simple comparisons.
   • Presents options with cost, risk and time.
   • Confirms the decision and the next steps.

1. You are designing a customer data platform for a Singapore retailer. How do you build in Personal Data Protection Act (PDPA) requirements?

   What a good answer covers

   • Identifies which fields are personal data and tags them in the model.
   • Limits use to the purposes the customer agreed to, and tracks consent.
   • Sets access controls, retention rules and a way to delete or correct records.
   • Plans how the organisation would detect and report a data breach.
   • Involves the data protection officer early.

2. How would the design of a data platform change if it serves a bank regulated by the Monetary Authority of Singapore (MAS)?

   What a good answer covers

   • Expects stricter rules on security, audit trails and where data and its backups are held, including with outsourced providers.
   • Adds strong access control, encryption and logging.
   • Considers rules on outsourcing and cloud use that apply to financial institutions.
   • Works with risk and compliance teams before choosing platforms.
   • Checks requirements against current MAS guidance.

Common questions about a data architect job

Sources