Data warehouse migrations are notorious for being complex, time-consuming, and error-prone. Converting SQL syntax, adapting to different function signatures, and ensuring data types map correctly across platforms can take weeks or even months of tedious work. Today, we're excited to demonstrate how DinoAI – our "cursor for data" – can dramatically accelerate this process.

Watch the full demonstration on YouTube.

The Challenge: Data Warehouse Migrations

When organizations decide to migrate from one data warehouse to another – whether for cost savings, performance improvements, or to take advantage of new features – they face several significant challenges:

1. SQL syntax differences: Each warehouse has its own dialect of SQL with unique functions, data types, and syntax requirements

2. Business logic preservation: Ensuring that transformed code maintains the exact same business logic

3. Testing and validation: Verifying that migrated models produce identical results

4. Documentation: Creating and maintaining reference guides for syntax differences

5. Scale: For large repositories, manual conversion becomes exponentially more complex

As Fabio Di Leta (Co founder, Paradime) noted during our livestream, even migrating just a handful of models can take "three to four hours at minimum" – and real-world migrations often involve hundreds of models.

Introducing DinoAI for Data Migrations

DinoAI transforms this process by leveraging multiple context sources to understand both your original code and the target syntax, then automatically handling the conversion while preserving business logic.

What makes this approach unique is that DinoAI isn't just a code translator – it's an intelligent assistant that combines:

• Multi-engine warehouse connections: Test and validate against both environments simultaneously

• Context support: Use migration guides as reference material for more accurate conversions

• Error-driven learning: DinoAI can consume error messages and adapt its approach

• .dinorules guidance: Provide specific instructions for handling the migration

Let's explore how this works in practice with a real migration from Redshift to Trino.

Setting Up for Multi-Engine Development

The first step in any migration is establishing connectivity to both environments. In Paradime, this is straightforward – you can configure connections to multiple warehouses and switch between them with a simple target flag:

This enables you to test the same model against both environments, identifying compatibility issues before starting the migration.

Leveraging Migration Guides as Additional Context

One of the most powerful features we demonstrated is DinoAI's ability to use migration guides as additional context. While we showcased a PDF format in our demo, the core benefit is providing comprehensive reference material to the AI agent, regardless of format.

Migration guides serve as crucial knowledge repositories that DinoAI can reference when making conversion decisions. Fabio explained that traditional AI assistants have limited context windows, restricting how much information you can provide in a single conversation. By adding comprehensive migration guides as context, DinoAI gains access to detailed reference material needed to handle complex conversions with greater accuracy.

In our demonstration, we used a guide that detailed:

• Data type mappings between Redshift and Trino

• Function equivalents (e.g., LIST_AGG → ARRAY_AGG)

• Syntax differences for common operations

• Best practices for the conversion process

What matters isn't the format, but that DinoAI can access and understand this enriched context to make more intelligent decisions. Interestingly, this guide was itself created with DinoAI's help – a "meta" approach that showcases the compounding benefits of AI assistance.

The Migration Process in Action

Our demonstration followed a systematic approach to migration:

1. Baseline Testing

We began by running a simple model against Redshift to establish baseline functionality:

This model ran successfully against Redshift, confirming our starting point.

2. Initial Compatibility Testing

Next, we ran the same model against Trino to identify potential compatibility issues. In this case, the simple model worked in both environments without modification – a common scenario for basic SELECT statements.

3. Advanced Model Migration

Moving to a more complex model with aggregations and window functions, we encountered our first errors when running against Trino. This is where DinoAI's power became evident.

By providing the SQL file, error message, and migration guide as context, DinoAI automatically:

1. Identified incompatible functions and syntax

2. Suggested replacements based on the migration guide

3. Made the necessary changes while preserving business logic

4. Added comments explaining the modifications

For example, DinoAI converted:

To:

DinoAI doesn't just convert the code silently—it automatically adds detailed comments to document the changes it made. After performing the SQL conversion, DinoAI inserts helpful comments like this at the top of the file:

4. Recursive Error Resolution

When our first migration attempt still produced errors, we simply copied the new error message back to DinoAI, which then refined its approach. Fabio described a streamlined workflow: when encountering issues, he simply copies the error message from the terminal and sends it directly to DinoAI, allowing the agent to diagnose the problem and suggest the appropriate fix without manual analysis.

5. Iceberg Format Configuration

When our first migration attempt still produced errors, we simply copied the new error message back to DinoAI, which then refined its approach. Fabio described a streamlined workflow: when encountering issues, he simply copies the error message from the terminal and sends it directly to DinoAI, allowing the agent to diagnose the problem and suggest the appropriate fix without manual analysis.

This enables true multi-engine analytics, where the same data can be accessed by multiple processing engines.

Real-World Impact

The approach we demonstrated doesn't just save time – it fundamentally transforms the migration experience:

Reduced migration time: What would take hours or days manually can be completed in minutes with AI assistance.

Higher accuracy: By systematically applying rules from migration guides rather than relying on human memory, the conversion is more consistent.

Better documentation: DinoAI automatically documents the changes made, creating an audit trail and reference for the team.

Progressive validation: Test each model in both environments as you convert, rather than waiting until the end.

Knowledge preservation: The migration guide becomes a persistent asset for future work.

Beyond Redshift to Trino

While our demonstration focused on Redshift to Trino migration, the same approach works for other common migrations:

• Snowflake to BigQuery

• BigQuery to Databricks

• Redshift to Snowflake

• Any other combination of supported engines

The key is that DinoAI understands both the source and target dialects through the context you provide.

Advanced Use Cases

Beyond simple migrations, this capability enables several advanced scenarios:

Multi-engine development: Maintain a single codebase that works across different warehouses.

Progressive migration: Move models gradually while maintaining a functioning system.

Iceberg adoption: Transition to modern table formats that enable true data mesh architectures.

Vendor flexibility: Reduce lock-in by ensuring your models can run in multiple environments.

Getting Started with DinoAI for Migrations

If you're planning a data warehouse migration, or even just want to explore the possibility of multi-engine compatibility, here are some steps to get started:

1. Set up connections to both environments in your Paradime account

2. Create or obtain a migration guide for your specific warehouse combination

3. Start with simple models to understand the basic patterns

4. Use DinoAI's context features to provide both the code and migration guide

5. Iterate progressively, fixing errors as they arise

The most successful migrations follow this incremental approach, building confidence and knowledge as you progress.

Try It Today

DinoAI's migration capabilities are available now in Paradime. If you're planning a data warehouse migration, considering multi-engine development, or exploring Iceberg format adoption, this approach can save your team weeks of tedious work while improving accuracy and documentation.

Current Paradime users can start using these features immediately. New users can try DinoAI for free today and see how it transforms your data engineering workflow.