AI automation for Indian businesses.
The repetitive work in your business, running unattended, tested, and with a visible trail of what it did. This is the thing I actually do for a living.
Fixed quote per workflow after a free 30-minute call. Most automations cost less than the salary of the hours they return.
Agreed as a date in writing before anything starts.
What this means in practice
Not a chatbot on your website. The work that eats a person's day: copying data between two systems that do not talk to each other, generating the same report every Monday, chasing invoices, sorting and routing incoming enquiries, publishing content on a schedule, keeping a spreadsheet and a database in sync.
Each of those is a workflow with a trigger, a set of steps and a definition of success. Once that is written down precisely, it can run without a person — and it will run at 3am on a Sunday, which the person will not.
Why I am credible on this specifically
I am an AI Engineer at Aaziko Global LLP. In August 2026, AI agents made 99.77% of my line changes; in October 2025 they made none. The whole of my working practice is now writing the specification, directing agents to implement it, and reviewing and testing what comes back. Automating other people's work is the same skill applied outward.
AI-PULSE is the public proof. It is a YouTube channel that publishes itself: one long-form video and one Short about a trending AI tool, every day, with nobody pressing go. It runs entirely inside GitHub Actions — no server, no cron box, no monthly bill — and its own tests run before the publish step, so a broken run fails loudly instead of publishing something wrong. That last detail is the difference between automation you can leave alone and automation you have to babysit.
The failure mode I design against
An unattended system's danger is not that it stops. It is that it keeps going and quietly does the wrong thing for three weeks before anyone notices. So every automation I build has three things: tests that gate the step which is expensive to get wrong, a log of what it did and when, and an alert that reaches a human when it fails.
Silence is cheaper than a wrong invoice. If the pipeline is not certain, it stops and tells you.
Cost, and how to tell whether it is worth it
Automations start at ₹30,000 per workflow and most take one to two weeks. The arithmetic is usually simple: count the hours a week the task currently takes, multiply by what that person's hour costs you, and multiply by fifty. If the automation costs less than that, it pays for itself inside a year and keeps paying every year after.
Where it is not worth it: tasks that happen rarely, tasks whose rules change every month, and tasks where a human judgement call is the actual work. I will tell you which of your processes are in that category rather than automating them badly.
Proof
Shipped, measured, inspectable.
AI-PULSE
A YouTube channel that publishes itself: one video and one Short about a trending AI tool, every day, unattended — on GitHub Actions at zero infrastructure cost.
The Observable Machine
The site you are reading — a portfolio that instruments itself: its own analytics pipeline, a public /stats page and an admin, all inspectable.
CGPE Connect
Replaced spreadsheet-and-email operations with realtime, role-separated workflows — live dashboards, an S3 document vault and an audit trail behind every action.
Questions
Asked and answered.
If yours is not here, ask it on the call. I reply within 24 hours.
What can actually be automated in my business?
+
Anything with a clear trigger and repeatable rules: moving data between systems, recurring reports, invoice and quote generation, routing enquiries, scheduled publishing, syncing a spreadsheet with a database, monitoring something and alerting when it changes. If you can write down the steps, it can usually run itself.
How much does an automation cost?
+
From ₹30,000 per workflow, quoted fixed after a free 30-minute call. Work out the hours a week the task takes now, multiply by the hourly cost and by fifty — if the automation costs less than that, it has paid for itself within a year.
Will it need a server I have to pay for every month?
+
Usually not. AI-PULSE — which publishes two videos a day, unattended — runs entirely inside GitHub Actions with no infrastructure cost at all. I default to that and only add a server when the work genuinely requires one.
What happens when it breaks?
+
It stops and tells you, rather than continuing and doing the wrong thing. Every workflow gets tests in front of the step that is expensive to get wrong, a run log you can inspect, and an alert that reaches a person. Silence is cheaper than a wrong invoice.
Is this the same as ChatGPT writing my emails?
+
No. That is a person using a tool. This is a system that runs on a schedule or a trigger, without anyone opening it, with tests and logging around it. An AI model may be one step inside the pipeline, or there may be no model at all — the value is in the workflow being reliable, not in which technology is inside it.
Do you take over automations someone else built?
+
Yes, and the first thing I do is find out whether it has any tests or logging. Most inherited automations have neither, which is why nobody trusts them. Adding those is often cheaper than rebuilding and is where I would start.
Open to full-time & freelance
Tell me the problem.
I'll tell you the price.
A free 30-minute call, no obligation. If I am not the right person for it, I will say so and point you at who is.