Every engagement starts small, proves itself with your own data, and only then grows
Four phases. Each has a duration, a deliverable and a decision at the end. You can stop at any gate.
- Phase 1
Understand
1 to 2 weeks
What happens
Site visit or working sessions with operations, not only IT. We map the process and the pain in your words.
You receive
A written problem statement and an inventory of the data you already have.
The gate
Agree the one problem worth solving first.
- Phase 2
Assess
2 weeks
What happens
We pull samples from your systems and check quality, gaps and whether the records can be linked. Quick baseline numbers.
You receive
A data readiness report with a go, fix-first or no-go verdict, and the pilot scope and price if it is a go.
The gate
Go, or a fix-first data plan. Sometimes the honest answer is no-go, and the report says so.
- Phase 3
Pilot
6 to 10 weeks
What happens
We build the first working system on one line, one branch or one department, with your people involved weekly.
You receive
A running system plus a measured before and after on the number agreed in phase one.
The gate
Scale, adjust, or stop.
- Phase 4
Deploy and support
Ongoing, monthly
What happens
Roll-out, training, monitoring, and retraining models as the business changes.
You receive
A monthly review and a support agreement your IT team can also act on.
The gate
A quarterly review of value delivered.
Three ways to engage
Readiness assessment
Fixed fee · 2 weeks
The assess phase on its own. You get the readiness report and a scoped, priced pilot proposal. Many clients buy only this first.
About the assessmentScoped pilot
Fixed scope · fixed price · 6 to 10 weeks
One problem, one unit, one measured result. The price is fixed in the pilot proposal, once the assessment shows how much data and integration work is involved.
Retained engineering
Monthly
After a pilot, or for businesses that want a data and ML team without hiring one. Agreed hours, a named lead, a monthly review.
Where it runs
On your own servers or your own cloud account wherever possible. Source code, pipelines and documentation are handed over. The tools are standard and well supported, so your own IT team can maintain the system, and nothing stops working if we part ways. Data does not leave India, or your infrastructure, without written agreement.
Three more rules for AI work
Models, assistants and agents add a kind of risk that a report does not: they can be confidently wrong. These rules apply to every engagement in the AI and ML engineering group.
- Rule 1
A test set before a model
A few hundred of your own real cases, with the right answer, are collected before anything is trained or prompted. Every change is scored against them, and the score is shared with you.
- Rule 2
A person at every irreversible step
Anything that moves money, changes a record for good or reaches a customer waits for a person's approval. Approval is removed one action at a time, as the test set earns it, never all at once.
- Rule 3
Cost and accuracy, published monthly
What a prediction, a conversation or a completed task costs, next to how often it was right. If either number slips, the monthly review says so, and says what we will do about it.
The harness is described in detail under AI agents and evaluation harnesses, and the monitoring under ML operations and monitoring.
What we will not do
- Re-implement your ERP, CRM or LMS. We work on top of what you have.
- Move your data outside India, or outside your own infrastructure, without written agreement.
- Sell you a model where a clear report or a simple rule would do the job.
- Start a pilot before the data assessment says the data can support it.
- Quote a timeline we have not delivered before on data like yours.
Who you work with
An engineering lead is on every project from the first visit to handover, with a data engineer and, where the problem needs it, a domain analyst who has run a plant floor, a service operation or an admissions office. Meet the team
Not sure if your data is ready?
Take the free 20-minute readiness check. You get a short scorecard and a plain answer about what, if anything, is worth doing next.