All research areas
Research
Artificial Intelligence
Applied AI systems built to survive contact with production.
We are interested in AI that ships. That means evaluation before deployment, a measurable baseline, and a clear account of how the system fails. A demo that works once is not a result.
In production
CMS — plagiarism & similarity screening
ML-backed similarity checks run inside CMS's submission pipeline — applied model work living in a product, not a notebook.
See the product →What this looks like in practice
- Model selection, fine-tuning, and evaluation harnesses
- Retrieval-augmented generation over private corpora
- Agentic workflows and tool-use architectures
- Inference optimisation and cost modelling
- Human-in-the-loop review and guardrails
Related publications
- Why the n mattersMd Mahmudul Hoque · 17 July 2026