Forward-deployed AI engineering training
Forward-deployed AI engineering training is for engineers and engineering leaders who need to build AI systems that run inside a real organisation, and who need a way to judge whether they can do it. The training is practical. Participants build against realistic cases, and each piece of work is assessed against a written standard.
I am Mayur Patil, based in Nashik, India. I run the public cohort and the workshops through AIXL Academy, and I deliver the same material to company teams.
What the training covers
The method follows the work of a forward-deployed engineer: discovery, architecture, integration, and handover. Each stage has a written output and a rubric. The four gates are the structure I use to assess that work, and the forward-deployed engineering page sets out the method in more detail.
Participants also learn the parts around a model that decide whether an agent is reliable. The harness engineering term describes those parts, and the evaluation harness term describes how the work is measured.
The public cohort
The public cohort is a live programme with a weekday-evening track and a weekend track. Participants build one realistic system across the course, and the assessment is about judgement at each gate, not about completing exercises. The programme and its registration are described on the AIXL Academy cohort page, which is the reference for dates, session times and terms. I do not repeat those details here, because they change.
The cohort is open to individuals. It suits engineers who want to work forward-deployed, and engineers who already do that work and want a structured way to judge their own performance.
Team programmes
A team programme is built on your team's tools, use cases and data, and runs for one enterprise team at a time. It is designed after a short conversation about the team's current work, and it ends with the team's own working system and an assessment of each person's demonstrated skills.
Team programmes are a good fit when a capability is needed quickly and the team has real work to bring. They are not a generic course delivered to a room of people who do not share a project.
FDE curriculum for firms that will deliver it
Some firms want to run the programme themselves, with their own trainers, for their own staff. For those firms I design the curriculum, the scenarios, the assessment rubric and the materials, and I train the trainers who will deliver them. The curriculum is built to be taught by people who did not write it, so it includes the assessment standard and the reasoning behind it.
Time zones and delivery
The public cohort runs live online. Session times are published by AIXL Academy, and the weekday-evening and weekend tracks are offered so that participants in different time zones can choose a track that fits their working week. For team programmes, the schedule is agreed with the team.
I work with teams in India, the UAE, Singapore, the United States, the United Kingdom and Europe. Country pages cover the rules that apply to the systems participants will build: UAE, United States, Singapore, Europe and India.
Who should not enrol
The cohort is not for someone who wants a certificate of attendance without doing the work. Participants are assessed, and the assessment is reported honestly. It is also not the right first step for a leader who needs delivery done for them. For that, the FDE consultant page describes the consulting options.
Global capability centres
Training for a capability centre usually runs as a team programme, so that the centre builds the same capability across several business units. The global capability centres page describes how that works for centres.
Why the full path matters
Most AI training stops at the tool. A participant learns to call a model, to write a prompt and to connect a tool. Those skills are useful, but they do not answer the questions a team meets in production: which data the agent may read, who approves an action that changes a record, and how the team knows the system still works after a change. Forward-deployed training covers those questions, because the work in a client environment is where they arise.
The programme follows the full path: discovery, architecture, integration and handover. The four gates are the structure I use to assess each stage, and each assessment is written against a rubric.
What is current in 2026
The governance material a participant must know is current. The IMDA Starter Kit for Testing Agentic Systems, version 1.0, released in October 2026, sets four steps: identify the relevant risks, create realistic tests, run the tests and analyse the logs, and monitor and re-test continuously. The MAS Guidelines on Artificial Intelligence Risk Management were finalised on 7 October 2026 and take effect on 7 October 2027. India's Digital Personal Data Protection Rules were notified on 13 November 2025, and most of their obligations apply from 13 May 2027. The programme covers these, with the sources.
The IMDA agentic testing kit and the MAS guidelines are glossary terms with the official wording.
How a team programme is scoped
A team programme starts with a short conversation about the team's work, its tools, the data it can reach and the workflow it wants to build. From that, I design the scenarios so that they match the team's systems, not a generic case. The programme ends with a working system the team owns, an evaluation against an agreed set, and a runbook for the named owner.
Each participant is assessed on what they demonstrated. The assessment is written, and it is shared with the participant and, where the team asks, with its manager.
What participants keep
A participant keeps four things. A discovery brief for a real workflow, an architecture decision record they wrote, an evaluation report with its test set, and a handover note. These are working documents from their own work, not workshop notes. A team keeps its system, its runbook and its evaluation set, which the team can run after I leave.
Where the programmes run
The public cohort runs live online, with a weekday-evening track and a Saturday and Sunday track. Dates and timings are on the AIXL Academy cohort page."
Team programmes run for teams in Singapore, Dubai, Bengaluru, Hyderabad and the Bay Area, in the team's own hours, with a short overlap for decisions. The place pages describe each market: Singapore, Dubai, Bengaluru.
City training pages cover Hyderabad, Mumbai and Pune, and the India page covers the public cohort and team programmes across the country.
Questions to settle before a team programme
Three questions settle most of the scope. Which tools does the team use, and which of them may read company data? Which workflow would the team be proud to ship in the next quarter? Who will own the system after handover? A team that can answer these three in the first conversation finishes its programme with a system in production.
Who the programme is for
The programme suits three groups. Individual engineers who want to move from building features to building systems that run inside a client environment. Engineering leads who must judge whether their team is ready to ship an agent. And consultants and trainers who want a structured standard for their own teams. Participants usually have a few years of software experience and some exposure to a model, but they do not need prior experience of agents in production.
How the gates work in practice
Each gate has a written output. The discovery gate produces a brief that states the outcome, the owner, the data and the evidence that matters. The architecture gate produces a decision record, with the options considered and the reasons for the choice. The integration gate produces a working build, tested against an evaluation set that the owner has seen. The handover gate produces a note, a runbook and a named owner. A participant passes a gate when the output meets the rubric, not when they have completed a set number of exercises.
What a rubric looks like
A rubric names the criteria for a gate and gives examples of each level. For the discovery gate, one criterion is whether the outcome is stated in a form the owner can check. Another is whether the data needed is named and its use agreed. Participants and assessors use the same rubric, and disagreements are discussed, so that the standard stays the same from one cohort to the next.
Who the programme is not for
The programme is not a fit for someone who wants a certificate without the work, or for a team that needs delivery done for it. Those teams are better served by the consulting hub, where a partner builds with them. The programme assumes the participant will do the work and be assessed on it.
How to start
For the public cohort, use the registration on the AIXL Academy page. For a team programme or a curriculum, send an enquiry with the team size, the tools in use and the use cases the team wants to build. I reply with a suggested format.
Common questions
Who is the public cohort for?
Engineers who want to work forward-deployed, and engineers who already do and want a structured way to judge their own work. Participants build a working system, not a set of exercises.
Is the cohort a bootcamp?
No. A bootcamp teaches tools. The cohort assesses judgement inside a realistic client environment: discovery, architecture, integration and handover, each against a written rubric.
What does a participant produce?
A discovery brief, an architecture decision record, a working build, an evaluation report and a handover note. I report what each person demonstrated, which is not the same as who attended.
Can a company run training for its own team?
Yes. Team programmes are built on your tools, your use cases and your data, and run for one enterprise team at a time. The exact format is set per programme.
Can you train our trainers to deliver the programme themselves?
Yes. For firms that will deliver the programme themselves, I design the curriculum, the scenarios and the assessment rubric, and train the trainers who will run it.
Where do the sessions run?
The public cohort is delivered live online through AIXL Academy, with a weekday-evening track and a weekend track. Session times and dates are on the AIXL Academy page, which is the reference for them.