Upcoming

The LLM Engineer certification is in production and not open yet. No date announced yet.

AI Application Engineering

LLM Engineer

Coming soon

Selects, adapts, serves and evaluates language models, and reasons about behaviour that changes between identical calls.

About this role

Placeholder — role overview not yet written

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.

What the exam will measure

A working outline, published early so the scope is arguable before it is fixed. It is not the final domain model, it carries no weights yet, and it will change as the competency standard for this role is written.

01Model selection and trade-offs
02Prompt and context design
03Adaptation: fine-tuning, distillation, LoRA
04Inference and serving
05Evaluating non-deterministic systems
06Retrieval and grounding
07Cost and latency engineering
08Failure modes and mitigations

How the exam will work

This part is already built. It is how every Abhyaas certification works, and it will not change for this one.

Generated from a blueprint: No exam is a random draw. Each certification declares its knowledge domains, their weights and its easy/medium/hard split as data, and the generator satisfies all of it exactly or refuses to build the paper.
60 scored questions: The standard objective form. Two candidates sitting the same certification get different papers built to the same rule, not the same paper twice.
Scored on our servers: Against a frozen snapshot of the questions exactly as they were served, so a credential earned years ago is still explicable.
A result that names the gap: A failed attempt returns a score per knowledge domain and the chapters that cover the weak ones. It is a diagnosis, not a verdict.
A credential a stranger can check: Every pass issues a credential with its own public verification page, showing the role, the level and whether it is still active.

Not decided yet

Named rather than left blank, because a missing price reads as a hidden one.

Levels: A certification may have two, three or four rungs — Foundation, Practitioner, Professional, Expert. How many this role gets depends on whether the work at each rung is genuinely different enough to assess separately. It is not decided.
Price: Not published. A certification is a one-time purchase with every retake its blueprint allows included; the figure for this role will be set when it opens.
Date: No date announced yet. A certification opens when its question bank can satisfy its own blueprint and has been reviewed — not on a date chosen in advance.

Upcoming

The LLM Engineer certification is in production and not open yet. No date announced yet.

AI Application Engineering

LLM Engineer

Coming soon

Selects, adapts, serves and evaluates language models, and reasons about behaviour that changes between identical calls.

About this role

Placeholder — role overview not yet written

What the exam will measure

01Model selection and trade-offs
02Prompt and context design
03Adaptation: fine-tuning, distillation, LoRA
04Inference and serving
05Evaluating non-deterministic systems
06Retrieval and grounding
07Cost and latency engineering
08Failure modes and mitigations

How the exam will work

This part is already built. It is how every Abhyaas certification works, and it will not change for this one.

Generated from a blueprint: No exam is a random draw. Each certification declares its knowledge domains, their weights and its easy/medium/hard split as data, and the generator satisfies all of it exactly or refuses to build the paper.
60 scored questions: The standard objective form. Two candidates sitting the same certification get different papers built to the same rule, not the same paper twice.
Scored on our servers: Against a frozen snapshot of the questions exactly as they were served, so a credential earned years ago is still explicable.
A result that names the gap: A failed attempt returns a score per knowledge domain and the chapters that cover the weak ones. It is a diagnosis, not a verdict.
A credential a stranger can check: Every pass issues a credential with its own public verification page, showing the role, the level and whether it is still active.

Not decided yet

Named rather than left blank, because a missing price reads as a hidden one.

Levels: A certification may have two, three or four rungs — Foundation, Practitioner, Professional, Expert. How many this role gets depends on whether the work at each rung is genuinely different enough to assess separately. It is not decided.
Price: Not published. A certification is a one-time purchase with every retake its blueprint allows included; the figure for this role will be set when it opens.
Date: No date announced yet. A certification opens when its question bank can satisfy its own blueprint and has been reviewed — not on a date chosen in advance.