Boutique

About Talorik

Software that reads operational data and returns a next step you can defend.

What we build

Talorik is a data analytics boutique. We build software for organizations that cannot wait for someone to write the right query. The product watches the feeds you already have, ranks what looks off, and proposes a Course of Action with a trail back to source.

Energy desks, supply chains, and other operations where delay is expensive. The desk spends the hour on the call.

What we do not automate

Most analytics tools make the report faster. Talorik takes on the exploratory phase. The system flags relevance and ranks options. Authority stays with the human. Every suggestion can be audited.

If you have to know the question before you can see the problem, you will miss the problem. That is the gap Inputless Analytics is built for.

The call

A ranked next step the desk can sign

Always reading

The model runs on the stream. You do not have to ask first

Stays on your network

Processing happens inside your perimeter. The model does not leave

How it is built

TypeScript on the edge, Python for the engines, Inputless DB for the graph.

TypeScript core

Eleven npm packages: tracker, dispatcher, cognitive SDK, context, storage, visualization

Python engines

Reasoning engine, Graph RAG, pattern models. The parts that score and explain a recommendation

Inputless DB

Twenty-two node types, thirty-plus relationship types. Ask in language; get a subgraph you can walk

Event routing

280-plus event types, routed by channel. Processing under 100 ms when the feed is live

How we ship

How the product is constrained.

Your network

The model runs where the data already lives. On-premise or private cloud. Nothing has to leave for the system to work. You keep the keys.

  • On-premise and private cloud
  • No required data export
  • You operate the stack
  • Access stays with you

Always on

The software does not wait for a report cycle. It keeps a live model of operations and flags deviations as they form. Someone still has to approve the call.

  • Continuous ingest
  • No query required to see a problem
  • Ranked options
  • Human sign-off

Existing feeds

Connect the sources you have. Operators keep their tools. You do not stand up a new BI layer to start.

  • Works off existing feeds
  • Fewer blank search boxes
  • Deploy without a new BI layer
  • Operators keep their tools

A trail

If a recommendation cannot be explained, it should not ship. Confidence, source nodes, and a record of what was shown to whom.

  • Plain-language rationale
  • Confidence on the call
  • Evidence back to source
  • Audit log

Accountability

If a call can move money, people, or kit, someone has to be able to reconstruct it.

Built to be reviewed

We do not treat governance as a policy PDF. The same path that produces a recommendation has to show how it got there: which records, which rules, which confidence.

Bias checks sit on the models. Lineage sits on the graph. Sign-off sits with the operator. If that chain breaks, the feature does not go out.

Fairness checks sit on the pattern models
A reasoning chain you can walk from alert to source record
A named owner for each automated suggestion

Data handling

Collect as little personal data as the job allows. Route with consent where it applies. Keep records inside the perimeter you agreed.

Model review

Pattern models are watched for drift and bias. A score without evidence is not a recommendation.

Who signed it

Each suggestion carries a rationale, a confidence figure, and the records behind it. No silent action.

After go-live

Install is the start of the work. The model has to stay honest as your operations change.

We stay on the account

We stay on the account after install. We help you tune sources, rules, and the ontology as the operation changes. Ordinary product work.

The model gets better as it sees more of your context: which alerts were real, which were noise, what the desk actually did. That history is yours.

On the account

Did the desk get a usable call?
We score the engagement on decisions made
The model ages with you
New plants, new counterparties, new rules: we retune rather than freeze a snapshot
You run it
You hold the environment. We can see what you allow us to see
History helps
Past calls and outcomes make later recommendations cheaper to trust

Where we operate

Intelligence for desks where delay is expensive

Energy

Planetary-scale operations and grid intelligence

Model generation, transmission, and market signals in one continuous view so teams see imbalance and risk before they cascade.

Energy
01/05
Explore all industries