Funded research datasets, and systems I built or supervised. Most of the Bangladesh work ran with undergraduate teams, which is how much of it was taught as well as built.
2022–23 · Co-Investigator · SUST Research Centre AS/2022/1/24
A speech corpus for research on recognition of spontaneous Bangladeshi Bangla, a setting where available corpora were mostly read speech.
2021–22 · Principal Investigator · SUST Research Centre AS/2021/1/19
Roughly 100,000 document–summary pairs, built to make encoder–decoder summarization trainable for a language where dataset scarcity was the binding constraint.
2020–21 · Principal Investigator · SUST Research Centre AS/2020/1/27
Around 50,000 sentences annotated with six basic emotions. Related work from the group appeared as EmoNoBa at AACL-IJCNLP 2022.
2019–20 · Principal Investigator · SUST Research Centre AS/2019/1/20
An examination platform handling fixed-answer questions alongside descriptive answers graded by a Python program.
2018–19 · Co-Investigator · SUST Research Centre AS/2018/1/11
A judge for running programming contests and programming examinations in several languages, for use across the university.
2017–19 · Team led by Prof. Muhammed Zafar Iqbal
Result processing for Bangladesh's national medical and dental college entrance tests — roughly 60 colleges and 100,000 candidates per cycle, where a processing error is a national news story.
2020–21 · Supervised across three student cohorts
Online examinations with multiple-choice, matching, coding and descriptive questions, program-based grading, and basic proctoring.
2020–21 · Supervised as a course project
Scheduling under scarce classrooms: allocate rooms and slots while respecting instructor preferences as far as the constraints allow.
2016 · Undergraduate thesis, supervised by Md. Eamin Rahman
A read-mapping tool for long, noisy Oxford Nanopore reads, using gapped minimizers and FM-indexing.