Last checked 2026-11-01

Forward-deployed AI engineering training for global capability centres

A global capability centre that keeps building technology needs its own people to build and run AI systems. Buying delivery each time creates a dependency that the centre cannot easily unwind. Training its own team makes the capability permanent, and it lets the centre take each system from discovery to handover without an outside team.

I am Mayur Patil, based in Nashik, India. I run team programmes for capability centres and train the trainers who deliver them, and I run the public cohort through AIXL Academy.

What the programme builds

The programme follows the work of a forward-deployed engineer: discovery, architecture, integration and handover. Each stage has a written output and a rubric. The team builds a working system on its own use case, and the assessment is about the decisions made at each stage, not about completing exercises.

The four gates are the structure I use to assess the work, and the forward-deployed engineering page sets out the method. Participants also learn the parts around a model that decide whether an agent is reliable: the harness engineering term describes them, and the evaluation harness term describes how the work is measured.

How a team programme is designed

I start with a short conversation about the team's current work, the tools it uses, the data it can reach and the use cases it wants to build. The programme is then designed for that team, so the scenarios are the team's own and not a generic case study. The conversation also decides which business unit the programme serves, because the rules and the owners differ between units.

The programme ends with the team's own working system, an evaluation of the system against an agreed test set, and a runbook that a named owner can take over. Each person's demonstrated skills are assessed and reported.

Running it across business units

A centre usually serves several business units with different rules, data and owners. The practical approach is to run one programme at a time, so that each unit's system is built to its own standards and handed to its own owner. The centre's central group can keep shared templates, the evaluation approach and the handover format, which keeps the systems comparable without making every project wait for an approval.

Working hours matter. A programme can run in the centre's working hours, with a short overlap for decisions with the business unit, and a written handover so that the unit's staff can run the system outside those hours.

Training the centre's trainers

Some centres want their own trainers to deliver the programme to the next team. For them I design the curriculum, the scenarios and the assessment rubric, and I train the trainers. The curriculum includes the reasoning behind the assessment standard, because trainers who understand the standard can apply it to cases the curriculum does not cover.

The centre keeps the materials and the assessment record. The aim is that the centre can run the programme without me.

Country rules that apply to the systems

The systems a team builds in a centre are subject to the rules of the markets where their data and users sit. The country pages summarise those rules and link to their sources: India covers the data protection law that applies to the centre's own processing, and United States, Europe, Singapore and the UAE cover the rules for the markets the centre's business units most often serve. The training uses these rules as part of the design, not as a lecture.

Relevant work

The organisations I have worked with and can name are on the about page. Among them are Tredence, Persistent and Walmart, and I have worked with Wipro and Infosys on capability programmes. I do not describe the detail of that work beyond what the clients have agreed to share.

Public option for individual engineers

Engineers in the centre who want the structured public route can join the cohort, which is open to individuals and runs live online. The registration and the session details are on the AIXL Academy cohort page.

Why a centre trains its own engineers

A centre that keeps building needs its own people to build and run systems. Buying delivery each time creates a dependency the centre cannot easily undo. Training its own engineers makes the capability permanent, and it lets the centre take each system from discovery to handover without an outside team. The centre also keeps the knowledge of why each system was built the way it was.

What the 2026 numbers mean for training

The Nasscom-Zinnov report released in July 2026 counts 2,117 GCCs in India, and coverage of the report says more than 1,200 of them have embedded AI and machine learning capabilities. A large share of those centres need engineers who can move from a model to a production system. The training is built for that gap.

What a team programme covers

A team programme follows the work of a forward-deployed engineer: discovery, architecture, integration and handover. The team builds a working system on its own use case. Each stage has a written output and a rubric, and the assessment is about the decisions made at each stage.

Three topics are always in scope for a centre. The first is data: which personal data an agent reads, under the Digital Personal Data Protection Rules, which apply most of their obligations from 13 May 2027. The second is permission: how a search respects the rights of the person asking, which is covered by the permission-aware retrieval term. The third is testing: the four steps of the IMDA Starter Kit for Testing Agentic Systems, which are useful for any centre whatever its market.

How the programme is run across business units

A centre usually serves several business units with different rules, data and owners. The practical approach is to run one programme at a time, so that each unit's system is built to its own standard and handed to its own owner. The centre's central group keeps shared templates, the evaluation approach and the handover format. That keeps the systems comparable without making every project wait for approval.

Training the centre's trainers

Some centres want their own trainers to deliver the programme to the next team. For them I design the curriculum, the scenarios and the assessment rubric, and I train the trainers. The curriculum includes the reasoning behind the standard, because trainers who understand the standard can apply it to cases the curriculum does not cover.

Where the training runs

Centre programmes run in Bengaluru and Hyderabad, and for centres that serve US business units, in US Pacific time where the overlap needs it. The place pages describe each: Bengaluru and the consulting page for global capability centres. I run the public cohort for individual engineers through AIXL Academy.

Who should join

A centre team programme works best with a mix of people. Engineers who will build the next system. A lead who will own it after handover. A reviewer from the business unit who will say whether the output is right. The programme is designed so that each role sees the part of the work it owns, and so that the owner sees how the system was built.

Measuring what a team learned

A team's learning is measured by what it ships. At the end of the programme, the team has a working system, an evaluation report with its test set, and a runbook. The evaluation report shows how the system performs on the agreed cases, and the team can rerun it. The runbook shows whether someone outside the team could operate the system. Those two documents are the measure, and they stay with the centre.

After the programme

A programme ends with a handover, but it does not end the centre's work. The centre should schedule a re-run of the evaluation each time a model, a prompt or a source system changes, and it should review the runbook after the first real incident. Many centres assign one of the participants as the standing owner of the evaluation set, which keeps the knowledge in the centre.

How to start

Send an enquiry with the centre's size, the business units it serves, the tools in use and the first use case the team wants to build. I reply with a suggested format and a start date agreed with the centre.

Send an enquiry

Common questions

Why train a capability centre's own team instead of buying delivery?

A centre that keeps building needs its own people who can build and run systems. Training makes the capability permanent, and the team takes its own work from discovery to handover.

What does a team programme for a centre include?

A short conversation about the team's current work, a programme built on the centre's tools, use cases and data, and an assessment of each person's demonstrated skills. The team finishes with a working system of its own.

Can the programme run across several business units?

Yes. Programmes are designed for one team at a time, and a centre can run several in sequence so that each business unit builds a system with its own standards.

Can you train the centre's trainers?

Yes. I design the curriculum, the scenarios and the assessment rubric, and train the people who will deliver the programme inside the centre.

Is there a public option for individual engineers in the centre?

Yes. The public cohort is open to individuals and runs live online. Its registration and session details are on the AIXL Academy page.

Enquiry

Tell me about the work

The use case, where it is blocked and who is involved. I reply with a suggested starting point. Or email hello@mayurpatil.ai.

Goes to my CRM and my inbox. No newsletter, no sharing. See Privacy (opens in a new tab).

Close