Platform

One intelligence layer over the data you already have.

ReefLink sits between your organizational data and the people who make decisions. It queries sources in place through the Trino distributed SQL engine, applies your security policies to every request, and adds search, investigation, AI analysis and monitoring on top.

Request a Demo View connectors

Architecture

Layers with clear responsibilities.

Sources stay where they are. The engine plans and distributes work. Governance applies to every layer above your data.

Interfacesfor people and tools

WorkspaceSearchEntity profilesMonitorBriefsSQL editorREST APIJDBC / BI tools

Intelligence servicescontext and analysis

AI query serviceMetadata & glossary contextEntity resolutionAnalyticsWatch rules

Distributed query engineTrino

Coordinator · plan & scheduleWorkers · parallel executionResource groupsIn-memory pipelines

Catalogs & connectors30+ source types

Custom Teradata connectorIceberg / Hive / DeltaPostgreSQLOracleKafkaSnowflake+ more

Your sourcesunchanged, in place

WarehousesOperational databasesData lakes & object storageStreamsSearch indices

Federation

One query. Many systems.

Standard SQL joins a Teradata warehouse, a lake on object storage and an operational database in a single statement. The coordinator builds a distributed plan; workers execute it in parallel.

Query / subscriber usage and complaints by segment3 catalogs · 5 workers

SQL

SELECT c.segment,
       count(*)           AS subscribers,
       sum(u.usage_gb)    AS usage_gb,
       count(x.ticket_id) AS complaints
FROM teradata.crm.customers c
JOIN lake.usage.daily u
  ON c.subscriber_id = u.subscriber_id
LEFT JOIN postgres.care.complaints x
  ON c.subscriber_id = x.subscriber_id
WHERE u.day >= DATE '2026-09-01'
GROUP BY c.segment

Distributed plan

Stage 0

Output
coordinator · final aggregate

Stage 1

Hash join + partial aggregate
5 workers · in memory

Stage 2

Scan teradata.crm.customers
projection pushedfilter pushed

Stage 3

Scan lake.usage.daily
26 of 365 partitionsParquet

Stage 4

Scan postgres.care.complaints
predicate pushed

Illustrative plan. Actual plans depend on connector capabilities, statistics and data layout — the proof of concept measures them on your data.

Teradata pushdown

Send the work to the data. Return the answer.

A generic JDBC connection pulls raw rows into the query engine. ReefLink's custom Teradata connector pushes supported operations into Teradata's AMP-parallel engine, so only results cross the network.

Standard JDBC connector

Moves raw rows, computes later

Data movedWork in source

Filters applied after transfer
Aggregation and joins in the query engine
Single session reads
Network and worker memory become the bottleneck

ReefLink Teradata connector

Computes in Teradata, returns results

Data movedWork in source

Predicates — WHERE, IN, BETWEEN
Aggregations — GROUP BY, SUM, COUNT, AVG
Joins resolved with Teradata's hash and merge strategies
Window functions — RANK, ROW_NUMBER, LAG, LEAD
FastExport for bulk parallel reads
Multiple sessions across AMPs

Bars are conceptual. Transfer reduction depends on the query and data; we baseline it during the proof of concept rather than promise a number.

Connectors

30+ source types, ready to connect.

Relational databases, warehouses, lakehouse formats, streams and object storage. Each is exposed as a catalog with its own permissions.

32 connectors

Relational databases

Cloud & lakehouse

NoSQL & streaming

Object storage

Interfaces

Everything in the browser.

No client installation. Analysts, administrators and security teams each get an interface built for their work.

For analysts

Query workspace

SQL editor with autocomplete and syntax highlighting
Natural-language questions to reviewed SQL
Schema browser across all catalogs
Paginated results with CSV and JSON export
Visual execution plan as a stage graph
Shared history, bookmarks and collaboration

For platform teams

Admin console

Cluster topology and worker health in real time
Inspect, kill and review queries with full plans
Connector and catalog management
Resource groups: CPU, memory, concurrency per team
Add or remove workers at runtime
Configuration history with rollback

For security teams

Policy management

Resource policies at every level
Row-filter and column-masking builders
Tag-based policies by data classification
Users and groups synced from LDAP / Active Directory
Audit log browser with filters
Policy promotion: dev → staging → production

Deployment

Runs where your data runs.

ReefLink deploys as containers in your data centre or private cloud. A coordinator plans queries; workers scale horizontally and can be added or removed without downtime.

Containerized
Coordinator, workers and services run as containers with protected secrets injection — no credentials in configuration files.
Scales horizontally
Add workers as concurrency and data volume grow. Resource groups keep teams from starving each other.
Your AI model choice
Use an approved managed model, or a self-hosted model on dedicated infrastructure for air-gapped environments.
Read-only by default
Source connections use read-only service accounts scoped to approved datasets.

Turn your data into intelligence.

See ReefLink run against your own sources — federated queries, governed access, and AI analysis that shows its evidence — inside your environment.

Request a Demo Talk to an Expert