AI Technology Stack & Capabilities | ONAQO

WE PICK BORING
WHERE BORING WINS.

A stack is not a shopping list. What follows is grouped by the job each layer does, with a plain rating of how mature our practice is in it — including the parts where we would describe ourselves as competent rather than expert.

MATURITY KEY ●●●  Production, many engagements ●●  Production, fewer engagements ●●  Research or prototype only
REFERENCE ARCHITECTURE SIMPLIFIED · ADAPTED PER ENGAGEMENT
TIER 01 · SOURCES Operational databasesPostgres · MySQL · Oracle DocumentsPDF · Office · scans Sensors & camerasstreams · images SaaS & APIsCRM · ERP · tickets TIER 02 · DATA PLANE Ingestion · contracts · validation · lineage · feature store Airflow · Kafka · Spark · dbt-style modelling · warehouse TIER 03 · INTELLIGENCE Retrievalhybrid search · rerankpermission filters Model servingrouting · cachinghosted & open-weight Orchestrationtools · approval gatesstate · retries TIER 04 · SURFACE In-product features Assistants Queues & alerts Dashboards SPANNING LAYER Evaluationgolden sets · scoring Observabilitytraces · logs · cost Drift monitoringinputs · outputs Access & auditwho saw what Registriesmodels · datasets Kill switchesper feature NOT OPTIONAL
THE STACK

Twelve layers, rated honestly.

LAYER {{ layer.num }} {{ layer.maturityLabel }}

{{ layer.title }}

{{ layer.body }}

WHAT WE USE
{{ t }}
HOW WE CHOOSE

{{ layer.choose }}

RESEARCH — NOT PRODUCTION WE WILL TELL YOU WHICH ONE YOU ARE LOOKING AT

Things we are studying and would not yet sell you.

{{ r.status }}

{{ r.title }}

{{ r.body }}

BRING YOUR
ARCHITECTURE.

If you already have a stack, we would rather work inside it than replace it. Send us a diagram and we will tell you where an AI layer fits and where it will fight your existing systems.