About Octelion
Octelion Data & AI Solutions builds data, AI and the product around them, from Bengaluru, for companies in the US and India. This page is what that means in practice and what we will not do.
In plain words
Octelion is a small engineering company. We build the data platform that makes numbers agree, the reports people believe, the AI assistants and agents that do a defined job safely, the workflows that run on their own, the monitoring that tells a person when something breaks, and the web app that puts it all in front of the people who need it. We hand it over running, documented, and in your accounts.
It is for founders, operations leads, heads of analytics, product managers and CTOs at small and mid-sized companies, mostly in the US and India: people who are busy, who check claims, and who would rather read a query than an adjective. Usually someone who has a specific thing that is slow, missing or risky, and who has been through at least one data project that did not end well.
There are things we refuse to do
- We do not quote a number we are not prepared to hold to.
- We do not put an AI agent in front of customers without evals, a safety check and a human approval step.
- We do not present an invented metric as a measurement.
- We do not build a prototype and leave.
- We do not use your data to sell our work; everything public on this site runs on public data.
- We do not take work where the data cannot be made trustworthy within the scope.
The last one matters most: the report on top of untrustworthy data would be a lie with a chart.
Five principles
Running systems, not notebooks
A notebook proves an idea. A system survives a holiday weekend, and the Tuesday after it when the file arrives late and the column has changed. We build the second thing, with the scheduling, the checks, the alerts and the runbook, and we hand that over. If something we built cannot run without us, it is not finished.
Show the work on public data
We cannot show you client systems, so we show the same methods on datasets anyone can download: the NYC taxi trip records, Microsoft's Contoso sample. The numbers are computed, the query is on the page, and if you get a different answer we want to hear it. A method you can check is worth more than a logo wall.
The safety and operations parts are the work
Most prototypes stop at the moment they look impressive. The classifier in front of the agent, the approval step, the cost ceiling, the audit trail, the alert that carries a run link, the quality gate between two jobs: these are not the finishing touches. They are the difference between a demo and a system, they are the parts that fail in production, and they are in every quote from the start.
One written number
After the scoping call you get the scope, what is in, what is out, and one number for all of it, in writing, before any work starts. It does not move unless the scope does, and a scope change is its own written line. You should never learn the cost of a decision after you have made it. We would rather lose a project on the call than win it with an estimate.
Plain words
We write the way we talk on a call: short sentences, specific nouns, numbers you can check. Technical vocabulary has its place, on the pages where it does work, explained in a clause. It never substitutes for saying what the thing does. If we cannot explain a design in a paragraph a product manager can read, the design is not finished.
Bengaluru, remote
Octelion is based in Bengaluru and works remotely. For clients in the US that means we are online during your East Coast mornings and can take calls in your evening when a decision needs it, and work lands overnight on your calendar; for clients in India it means the whole working day. Weekly demos happen in your time zone. Everything we build lives in your cloud accounts, so where we sit changes nothing about where your data is.
One email is enough to start.
Tell us what is slow, missing or risky. We reply within one working day.