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.
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
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