Data layer
One version of the truth.
Pipelines and scheduled jobs that reconcile your systems, and a knowledge graph that tells AI what the records mean.
Three systems, three answers. Make it one.
What gets built
- Lakehouse pipelines
- Scheduled jobs and alerts
- API, SFTP, and webhook integrations
- Knowledge graphs
- Ontology design
- Feed reconciliation
- Postgres data models
- Data quality checks
- Dashboards
Built with Databricks, Asset Bundles, Python, SQL, Postgres, Prisma
Desk lamp DL-204, units in stock
Demo data
Storefront
41
Warehouse
38
Vendor feed
45
- Land
- Clean
- Match
Three systems, three answers
Teach your data what it means.
A knowledge graph links your records into one map of the business. Assistants follow the map instead of guessing.
Ask across the graph
Demo data
Select a circle to read its facts, or ask a question.
Jobs that run on time and say so when they fail.
Dry run first
Checked against real data before anything is written.
Alerts, not surprises
A failed step stops the run and tells you.
Yours to keep
Runs in your accounts, lives in your repositories.
Nightly inventory sync
Demo data
- Fetch vendor feedsnot started
- Clean and matchnot started
- Push to storefrontnot started
- Send summarynot started
Data builds
InfrastructureData
Warehouse inventory app
Replaced the spreadsheet. Barcode scanning, stock moves, and two-way storefront sync.
Next.js, Postgres, Vercel
Data
Vendor-to-storefront pipeline
Supplier feeds land overnight. Availability is correct online before the first order.
Databricks, Python
AIData
Inventory assistant
Staff ask ChatGPT what is in stock and get the answer from the system of record.
MCP, TypeScript
AIData
Demand forecasting
Every product shows where stock is heading, not only where it is.
Databricks, Python
Got a number nobody agrees on?
Thirty minutes to trace where it splits and what would fix it.