In development, entering pilot
AI-Examiner
AI-assisted evaluation for examinations at national scale. Built so that accuracy comes before speed, anonymity holds throughout, and every mark can be traced back to how it was given.
The problem
The largest manual task in the education system
Public examinations in Bangladesh generate answer scripts in the millions. Every one of them is read and marked by a human examiner, under time pressure, and a portion is re-checked afterwards. The system works, and has worked for decades, but its constraints are real and well known: the evaluation window is long, consistency between examiners is difficult to guarantee, and the cost of a single error falls entirely on one student's future.
This is the problem AI-Examiner is built for. Not to replace the examiner, but to give examination authorities a system where machine assistance handles extraction and consistency checking while human judgement stays where judgement is required. Solving this changes what a national education system is capable of, for every student in the country.
The design
How it is designed
AI-assisted handwriting extraction
Scanned answer scripts are processed to extract handwritten responses into structured, readable text, so evaluation is no longer limited by legibility or by the physical movement of paper.
Anonymised marking pipeline
Scripts are separated from student identity before evaluation and re-joined only after marks are finalised. Anonymity is enforced by the architecture, not by procedure.
Accuracy as the first priority
The system is designed around an explicit hierarchy: accuracy first, then efficiency, then speed. Where those goals conflict, accuracy wins by design. A faster system that is occasionally wrong is not an acceptable trade in examination evaluation.
Human examiners in the loop
AI-Examiner assists evaluation; it does not quietly decide results. Examiners review, adjust and confirm, with the system surfacing inconsistencies and outliers for attention rather than resolving them on its own.
Complete auditability
Every mark carries its record: what was extracted, what was suggested, who confirmed it, and when. Results are defensible under scrutiny because scrutiny was assumed from the start.
Built for scale
Architected for national-level volume from the beginning, rather than as a small system that would need rebuilding at the scale that matters.
Status
Status, and a straight answer
AI-Examiner is in development and entering pilot release. It has not been run on a public examination, and we will not suggest otherwise. What exists is a rigorously specified system, a designed architecture and a working development track.
We are looking for examination authorities and institutions who want to evaluate it properly, on controlled data, before anyone's results depend on it.
Ask us hard questions
We think this is the most important thing we are building. We also think examination integrity is the last place anyone should accept a marketing claim in place of evidence. Talk to us, ask hard questions, and test it yourself.
Straight answers
Questions about AI-Examiner
What is AI-Examiner?
AI-Examiner is an AI-assisted examination evaluation system built by ShikkhaNet Technologies for large-scale examinations in Bangladesh. It extracts handwritten answers from scanned scripts, runs an anonymised marking pipeline, keeps human examiners in the decision, and records a complete audit trail behind every mark.
Has AI-Examiner been used on a public examination?
No. AI-Examiner is in development and entering pilot release. It has not been deployed on a public examination, and ShikkhaNet states this plainly rather than implying otherwise.
Does AI-Examiner replace human examiners?
No. The system assists evaluation by handling extraction and consistency checking and by surfacing inconsistencies and outliers. Examiners review, adjust and confirm marks, so human judgement remains where judgement is required.
How does AI-Examiner protect student anonymity?
Student identity is separated from the answer script before evaluation and re-joined only after marks are finalised. Anonymity is enforced by the system architecture rather than by procedure.
What happens when accuracy and speed conflict?
Accuracy wins. AI-Examiner is designed around an explicit priority hierarchy of accuracy first, then efficiency, then speed, because a faster system that is occasionally wrong is not an acceptable trade in examination evaluation.
Can an examination board pilot AI-Examiner?
Yes. ShikkhaNet is seeking examination authorities and institutions willing to evaluate the system on controlled data before any live results depend on it. Contact the team to arrange a technical discussion.
Evaluate it before anything depends on it.
That is the right order, and it is the conversation we are asking for. Our team will walk you through the architecture in detail.