Writing & Research

• Submitted the under-five mortality paper to BMC Public Health (updated version; arXiv v2 now live). Extending the Depth Delusion scaling results with empirical validation on additional architectures.
• Preparing a comprehensive literature review on efficient transformers for low-resource settings as part of my PhD application reading list.
• Running experiments on multi-objective NAS using smaller search budgets suitable for single-GPU setups.

PhD Applications

Actively researching labs for Fall 2027 entry. I'm most interested in groups working on efficient ML, reliable evaluation of LLMs, or causal methods with development-sector applications. If your lab fits and you're reading this, feel free to email me.

Reading

• Causal Inference: The Mixtape by Scott Cunningham. Working through the instrumental variables and difference-in-differences chapters.
• The Alignment Problem by Brian Christian. Background reading on AI safety and value alignment.
• Papers: currently digesting FlashAttention-3, MegaByte, and recent ICLR 2025 efficient transformers track.

Teaching & Mentoring

Teaching data analysis and research methodology through Stat.BD to students, clinicians and other professionals, many of whom are doing their own research. Open slides and class notes are on the teaching page. Also informally advising two undergraduate students at Rajshahi on ML project methodology and research writing.

Deliberately not doing

• Taking on new Fiverr projects while finishing the papers.
• Starting any new research threads until the current two are submitted.