Home › Book a call
Book your SibylData technical call
Twenty minutes. One domain. You leave with your atomic unit defined, a first event schema, and the single benchmark that will tell you whether your retrieval is getting better.
Book my technical call → Email us instead
No calendar gymnastics: pick a slot, bring one real domain, and we talk. No obligation.
What happens on the call
The 20-minute agenda
We name your atomic unit
The one event that is your chunk boundary.
We sketch your core event schema
The typed fields and timestamp you lock in week one.
We define your benchmark queries
The shape of the 50-query set that guides every decision.
You decide
Whether to build it with us, or take the plan and run.
No pressure, no obligation
You keep the plan either way
The call earns its own time
If we are not the right fit, you still walk away with a usable data model and a clear next step. The call is built to be worth your time on its own. That is the deal.
Questions before you book
How much does this cost?
Every engagement is scoped to your domain and goals, so you get an exact figure on the call. The real question is what another quarter of unexplained retrieval regressions is already costing you.
Isn't this just another vector database?
No. A vector database stores embeddings. We define the data model, the context assembly layer, and the benchmark around it, so retrieval is deterministic and you can prove it works.
What if I miss an important event type in month one?
You will, and the structured catch-all catches it. You promote recurring patterns into first-class types each quarter, on a schedule, instead of discovering the gap in production.
We already ship with unstructured RAG. Why change now?
Because unstructured quality degrades as the corpus grows, and the rebuild only gets more expensive with more data. Day-one structure is the cheap version of work you will otherwise do later under live load.
How do you prove the architecture works?
With a number. We build a 50-query benchmark with known answers and gate every change against it, so improvement is measured, not claimed.
I'm not sure I can define a schema myself.
That is what the call is for. You bring the domain knowledge, we bring the schema discipline, and we sketch it together in the first ten minutes.
What about embedding model upgrades later?
Handled by design. We tag every chunk with its model version and run parallel indexes during transitions, so upgrades stay routine instead of becoming a rebuild.
Bring one domain. Leave with a data model.
Twenty minutes, one real domain, and a plan you keep whatever you decide next.
Book my technical call →