The platforms we build on and what we actually use each one for. Listed as capability rather than as badges, because we would rather tell you what we do with a tool than show you its logo.
Most engagements land on infrastructure the client already runs. We fit into it rather than asking you to move.
| Platform | What we build on it | Where it is the right choice |
|---|---|---|
| AWS | Bedrock for managed inference, OpenSearch or pgvector for retrieval, Lambda and ECS for the application layer, Step Functions for long running workflows. | The default when a client is already on AWS and wants inference, data and application inside one account boundary. |
| Microsoft Azure | Azure OpenAI for inference under enterprise terms, AI Search for retrieval, Functions and Container Apps for serving, Entra ID for identity. | Strongest fit where the content lives in Microsoft 365 and permissions have to be inherited from SharePoint and Entra rather than rebuilt. |
| Google Cloud | Vertex AI for inference and tuning, BigQuery as the analytics substrate, Cloud Run for serving. | Where the analytical data already sits in BigQuery and the AI layer should sit next to it rather than copy from it. |
| Databricks | Pipelines and transformation, feature and evaluation datasets, model serving, Unity Catalog for lineage and access control. | When the organisation already runs a lakehouse and wants AI workloads governed by the same catalogue as everything else. |
| Snowflake | The warehouse as the source of truth for structured retrieval, with governed access to the tables an agent is allowed to read. | Where the answer to a customer question is a query against warehouse data rather than a passage in a document. |
The integration layer is usually where the timeline actually goes. These are the systems we most often read from and write back to.
We do not currently hold formal partner status with the cloud providers, and we are not going to put logos on this page implying that we do. Where a client needs a badged supplier for procurement reasons, we will say so early. Where they need engineers who have built on the platform, that is what this page is describing.
Building on a platform and being a certified partner are different claims. We are making the first one.
Tell us what your stack looks like and we will tell you where an AI layer fits and where it does not.