اے آئی کی اصل انجینئرنگ
Everyone is learning to prompt. Almost nobody is learning to engineer the machine around the model — the harness, the loop, the graph, the retrieval, the system. Four weeks. One real project. Shipped live.
60 of 60 seats left · reviewed on a rolling basis
Pakistan has a very large number of capable programmers and a very small number who have ever designed an agentic system end to end. That gap is where the salaries, the contracts and the companies are. This programme is aimed squarely at it.
Learning to prompt
Learning to engineer the machine that prompts
Chaining calls in a notebook
Designing loops, graphs, budgets and stopping conditions
Treating the model as magic
Knowing what training made it good at, and what it never will be
A demo that works once
A system that survives real users, real cost and real failure
Each week ends with something that runs. Nothing here is theory for its own sake — the order exists because week four is impossible without week one.
Most people trying to build with AI fall over on the boring parts — types, boundaries, state, failure. Week one rebuilds those muscles properly, in TypeScript, with the standards a production team would actually hold you to.
A small service with a typed core, validated edges, and tests that fail for the right reasons.
DigitalPakistan already runs the ecosystem directory, the talent network, the incubation programmes and the investor side. Finishing this course plugs your work into all four of them.
Not a certificate. A working system published on DigitalPakistan, with your name on it and a URL you can send to anyone.
The strongest builds get put in front of the companies hiring through our talent network — with the code as the CV.
If what you build looks like a company rather than a project, it goes to the incubation and investment side of the platform.
Graduates get a listed builder profile in the ecosystem directory, tied to a verified Pakistan Pass identity.
To be plain about it: a job and investment are outcomes we work towards, not promises we sell. What is guaranteed is the teaching, the project, and that your finished work gets published and seen.
Funded and paying participants sit in exactly the same room, do exactly the same work, and nobody in the cohort is told which is which.
One payment, four weeks, everything included.
15 seats per cohort, funded by DigitalPakistan.
A GitHub link is the single most useful thing you can put in the application. It does not need to be impressive — it needs to be real.
Six minutes, no deck, no payment. Applications are reviewed on a rolling basis and read by a person, not a filter.
A four-week, project-based AI engineering programme from DigitalPakistan.io. It covers core software engineering, harness and loop and graph engineering for agentic systems, pretraining intuition and production RAG, and the art of system design. Every participant ships a real project that goes live on DigitalPakistan.
The programme is PKR 50,000. For this cohort it is reduced to PKR 12,000 for the full four weeks — the largest discount we will run. 15 fully-funded seats are reserved for applicants who genuinely cannot afford the fee.
Then you apply for a funded seat. It is free, there are 15 of them per cohort, and you do not need to submit documents or proof of income — you write a few honest sentences about your situation and a person reads them. Needing a scholarship has never counted against an applicant.
Yes. You should be comfortable writing and debugging code in at least one language, and comfortable reading code you did not write. You do not need machine learning experience, a computer science degree, or any prior work with AI systems.
A working agent harness of your own design — multi-tool, multi-step, with a graph you can defend — plus a grounded RAG system over a real corpus. In the final week both feed into a single system you design, ship and present, which is published on DigitalPakistan.
The start date is confirmed once the cohort is filled, and we give everyone at least two weeks of notice. Expect live sessions plus substantial project work each week. It is dense by design — four weeks is short, and the point is that you build something real inside it, not that you watch videos.
It can, and that is the point of shipping on the platform rather than into a folder. The strongest projects are put in front of companies hiring through our talent network, and projects that look like companies are routed to the incubation and investment side of DigitalPakistan. Nothing is guaranteed — the work decides.
No closing date is fixed yet — applications are reviewed as they arrive and the cohort closes when the seats are gone. Seats are limited, so earlier applications get more of our attention and more time to prepare.
Not at application. You only pay once you have received and accepted an offer of a seat. If you are offered a funded seat, you pay nothing at all.
Master AI Harness — PKR 12,000 for 4 weeks
60 of 60 seats left · reviewed on a rolling basis