Work
We would rather rule things out than sell you all of them.
Four practice areas. Most clients need one or two, and the honest answer is often that the third is not worth doing yet.
Implementation examples
Workflows with explicit human decision points
Manufacturing and order operations
Configurable-product order intake with human confirmation
Structured product requests, submission notifications, and a human review step before confirmation. The example describes verified test behavior without customer identities or unmeasured outcome claims.
Explore the work →Meeting operations and follow-through
Meeting intelligence with accountable follow-through
Transcript ingestion, structured action extraction, executive classification, and optional task handoff. The example describes implemented behavior without exposing participants, organizations, or meeting content.
Explore the work →Ongoing project
API cost control for an AI-enabled platform
Requirements and architecture for metering, spend limits, evaluated model routing, customer-owned credentials, and partner-agency delivery under one change-control process. No customer, location, or savings claim is published.
Explore the work →Advisory
Where AI is worth the effort, and where it isn't.
- Claude
- Databricks
- Snowflake
- dbt
- Python
Building
Assistants, automations, and internal tools that hold up in production.
- Claude
- Model Context Protocol
- TypeScript
- Python
- Postgres
- AWS
Data foundations
Warehouses, pipelines, and models — the part everyone skips.
- Databricks
- Snowflake
- dbt
- Airflow
- Python
- SQL
Staying with it
Monitoring, tuning, and support after launch.
- Evaluation harness
- Tracing
- Plausible
- Your cloud account
Not sure which one you need?
That is what the assessment is for. Describe the problem and we will tell you which practice it belongs to — or that it belongs to none of them.