DataQubeDataQube
Air-gap capable • Your infrastructure • No data leaves your environment

The AI data analyst that runs where your data lives

The organizations with the most valuable data are often the least able to use modern AI tools. DataQube delivers auditable analysis, reports, and boardroom decks entirely within your infrastructure.

On-premisePrivate cloudEvery answer traceable to its source
DataQubeAgent threadsAll systems operational
Why did we underperform the benchmark this quarter?
warehouse.sqlholdings + benchmark, as-of joinedevt-8f3a21
risk.engineBrinson–Fachler attributionevt-8f3a22
files.searchIMA §4.2 — single-issuer limitevt-8f3a24
Q3 Performance Review — IC · v23
compiled from this analysis · every figure traceable
Continue the analysis…

Used by teams across asset management, insurance, and banking.

01The difference

Don't take the data to the assistant. Bring the assistant to the data.

Powerful AI. Regulated analysis. No tradeoff.DataQube runs inside your infrastructure, where every answer is traceable, governed, and audit-ready.

Powerful AI. Wrong architecture.

*pastes holdings into the prompt*
Here's a confident answer.
provenance: unavailable · memory: this tab · data residency: elsewhere
  • Data must leave your environment
  • Answers are difficult to verify
  • Context disappears between sessions
  • Analysis lives outside your workflow

An analysis system inside your perimeter

warehouse.sqlholdings, as-of joined · 1,842 rowsevt-8f3a21
memory.recallbenchmark T+1 lag · validatedevt-8f3a24
provenance: attached · memory: the team's · data residency: yours
  • Runs on your infrastructure
  • Every answer links back to its evidence
  • Institutional knowledge is retained and reused
  • Analysis becomes reports and presentations

02One surface

The conversation is the notebook is the deck

Today the same analysis is recreated across chats, notebooks, spreadsheets, presentations, and tribal knowledge - copies that drift apart. In DataQube there is one record; chat, notebook, and presentation are just ways of reading it. They can't contradict each other, because there is nothing separate to contradict.

01

Ask

Ask questions in plain language. DataQube automatically runs the appropriate queries, models, and searches.

02

Verify

Inspect every result down to the underlying query, file section, or calculation.

03

Present

Generate reports and presentations directly from the analysis—with every figure linked back to its source.

Nothing changes silently - every edit is visible, history survives

03Platform

An operating system for enterprise analysis

Most organizations already have tools for data, documents, reporting, and governance. What they lack is a system that connects them into a single analytical workflow.DataQube serves as that system—bringing analysis, institutional knowledge, automation, presentations, and governance together in one place.

draft v24
v23 published · untouched

Presentations

Decks compile from analysis, not alongside it. Co-edit slide by slide with DataQube, every number traceable, and the version your committee received stays untouched until you publish the next.

Weekly IC PackMon 06:00
Quarterly ESG ScreenFri 07:00
ALCO Stress Pack1st of month
run #38 · parameters pinned✓ matches run #37

Scheduled Agents

Turn any analysis into a workflow - reproducible and deterministic, because nobody wants surprises in a key report - that re-runs on a schedule: the Monday IC pack rebuilds itself, flags what changed, and stages a draft for approval.

Drill into the Industrials drag…
warehouse.sqlas-of joined✓ evt-8f3a21
/pytop = drilldown(att, level="issuer")

Threads & Notebooks

Ask in plain language and get analysis you can defend - the queries, code, and reasoning attached. Your quants can verify any step in place instead of rebuilding it in a spreadsheet.

benchmark T+1 lag · validated

Institutional Memory

The conventions that make analysis correct usually live in heads and chat history. Here they're captured, validated, and recalled in every relevant run - and outdated knowledge is retired, never silently reused.

DataQube Data Gateway14 toolsfirst-party
Market Data9 toolsOAuth
Risk Engine6 toolsworkload id

Tool Catalog

Connect databases, file stores, and market data once at the org level. Credentials never reach the AI model, teams see exactly the data they're granted, and revoking access works instantly.

evt-8f3a21query.execute1,842 rows
evt-a10441permission.grantesg.api · once
evt-77d0e3auth.decisionwall-crossing✗ denied
evt-b81f55agent.runweekly-ic-pack

Audit & Permissions

When the regulator asks who saw what, the answer is an export, not a project. Permissions people understand, denials that always win, and a record of every query, file read, and refusal.

0

bytes of data sent back to us - in any tier

0%

of figures traceable to the query, rows, and file section behind them

0

isolated environment per analyst session - nobody shares anything

0 min

from install to the first fully audited answer

04Security & compliance

The part your CISO reads first

DataQube exists because regulated institutions can't send data to someone else's cloud. Every design decision starts there - security isn't a feature we added later, it's the constraint we built around.

Air-gap capableDeny always winsDatabases read-only at the source
custodydeploys into your cluster; the model runs on your GPUs
egresszero - no telemetry, no diagnostics, no license callbacks
isolationevery session in its own container, reclaimed after
authorizationexplicit deny always wins - barriers can't be overridden
data accessread-only connections, row limits, timeouts - at the database
evidenceevery query, file read, and denial on the record, SIEM-ready
licensinga signed file you control - nothing to call home to

05Deployment

Runs in your world, not ours

Most AI platforms require organizations to adapt to the vendor's infrastructure, security model, and deployment constraints. DataQube runs inside your environment - whether that's on-premise, private cloud, or fully air-gapped - so you retain control of your data, security, and operations.

Most common

On-premise Kubernetes

Helm charts into your cluster, images through your registry, logs to your SIEM. The same stack your platform team already operates.

AKS · EKS · GKE

Private cloud

Your tenancy, your VPC, customer-managed keys end to end. No shared control plane - there is nothing of ours to share.

Sovereign tier

Air-gapped

Built for networks that never touch the internet. Every feature works with zero egress - delivery, licensing, and support included. No exceptions to negotiate.

Trust is not an upsell - every tier ships the full security architecture.

From the field

Our auditors had questions; now they have provenance. The deck our investment committee reads on Monday built itself at 06:00 - and every figure links back to the exact query it came from.
IRHead of Investment Risk · Large Asset Manager
Deploys in your cluster - nothing leaves

See your data answer questions - without leaving your infrastructure

Thirty minutes with an engineer: live product, deployment options, and your security team's questions answered by someone who wrote the code.