What I am working on

For eighteen years my work has lived in other people's systems: a publisher's paywall, a university's faculty page, a professional network that decides who sees what. This site is the first place all of it sits together, and the first place where a student can search it without an account.

Thirty papers now have their own pages, with the citation metadata that lets Google Scholar index them properly, a copy-ready citation in two formats, and links to the publisher, the dataset or the code wherever those exist. There is a tool that answers questions from the publication record and shows you which papers the answer came from. And there are five guides, one per research thread, written for someone meeting the topic for the first time rather than for a reviewer.

Four papers went out this year, two of them on what AI costs and what it leaks. Eco-Transformers measures the energy cost of inference in large language models rather than only their accuracy, which is a question the field asks far too rarely. The sequential recommendation work moves my doctoral question onto models where the order of what you did matters, not just the set — and order turns out to be more identifying than people expect.

The teaching year was busier than the research one. The ACM Student Chapter at PSG iTech opened in October with 120 students, and in January we ran a session on digital footprints and privacy for 800. Standing in front of a room that size and explaining why deleting your name from a dataset does not make it anonymous is, I think, worth more than another conference paper on the same subject.

What is next: an ORCID, which is the one identifier I still do not have and which more journals now want at submission; course pages, once I have decided what of my own material I am willing to publish; and the working group on refute-problem difficulty for Indian computing learners, which I am co-leading into ACM COMPUTE 2026.

If you found something here useful, or wrong, write to me. That is the whole point of owning the channel.