2. Projects

I have made a number of things over the years. Some were archives. Some were tools. A few began as collaborations and others as open questions I never quite closed. Most are still online. A handful have disappeared entirely. Each one changed how I think about knowledge and form, usually in ways I only understood well after the fact.

Bichitra was the first large project I worked on. It is a digital variorum of Rabindranath Tagore's works, bringing together manuscripts, printed editions, editorial judgments and machine-readable metadata. You can read about the ambition of the project and the ingenious ways it was coordinated here. Bichitra remains one of the few projects I have been part of that feels finished, even now.

Recalling Jewish Calcutta was different. It was fragile from the start, and more personal too. Built with members of Calcutta's dwindling Jewish community, it held photographs and oral histories, alongside recipes and family trees that mapped a diaspora quietly disappearing. For a few years the site travelled widely, picked up by Time and the BBC among others. Then it vanished. A malware attack hit in 2017 and we recovered. Another hit in 2019 and we did not. The files survived. The site did not. That project taught me that recovery is not guaranteed just because you've managed it once. I wrote about what that taught me here.

I joined Letters of 1916 after those two projects. Like Bichitra, it was an editorial project, but the audience was different. Bichitra was built for scholars who already knew how to read manuscripts closely. Letters was built for the public. People I would never meet transcribed wartime correspondence from their own homes, on browsers I never got to test. My own work centred on the transcription interface and the network graphs that mapped how individual letters connected to each other. The editorial labour was distributed across hundreds of volunteers, and much of it is still active today. We wrote about the community participating in the project here. It was the first time I watched scholarship happen across that many hands at once.

Battle of Mount Street Bridge was public history built for a screen rather than a page. Built inside a game engine, it reconstructed a single day of the Easter Rising through a virtual city and a physics engine. We consulted military historians to test weapon accuracy and worked with mathematicians and geographers on the spatial modelling. External 3D artists built the environment itself. I contributed to the conceptual frame and the storytelling, working out what we were trying to show and what we were willing to simulate. It turned out to be a genuinely effective way of teaching the Rising to schoolchildren.

Humanities at Scale was about teaching, not in a single classroom but across an entire training platform. We built the content and delivery system from scratch, along with the workflows needed for multilingual collaboration between teams that had never worked together before. The project targeted teachers who were already doing the work but needed a framework to teach with. It was part of DARIAH-EU and brought together digital humanities researchers from across its working groups, coordinated over video calls and shared documents rather than any single office. In many ways it was a response to the crisis in the humanities. Looking back, I am not sure I ever believed in distance learning. At the time I was committed to the work and uncertain about the form. The effort of building something teachable and shared still matters to me.

anvay is a web-based topic modelling tool for Bengali texts. It began as a pedagogical experiment testing how to design digital tools for people who are not already experts. I wanted to avoid a structure built only for measurable outcomes. anvay was built with that openness in mind. The interface is exploratory and the documentation is interactive. You decide how deep to go, and reading what the model returns is left entirely up to you. You can use it with a single click here. It now works for a few other languages too, though I should warn you that these newly added languages are still experimental.

gridOCR is a desktop tool built for digitising historical printed books and periodicals, the kind full of running headers, footnotes and text that bleeds through ageing paper because nobody in 1887 thought about future archivists. It runs entirely locally. Give it scans and it splits double-page spreads and detects text regions. It then suppresses bleed-through and runs Tesseract OCR, before letting you correct individual regions and export clean plain text. It handles multiple scripts, though I honestly do not remember which ones work well. You will have to test it yourself. It grew out of a specific frustration. Most existing tools for this kind of work assume clean modern typography or require a cloud subscription, and neither is much use if you are working with nineteenth-century Bengali periodicals on a slow connection. gridOCR is not a general-purpose OCR tool. It is built for one kind of source and one kind of researcher, and there are probably about twelve of us.

overdub is a browser-based multitrack recorder. You open it, set a tempo and pick a key. Then you record. There is no subscription and no account required. I skip two-factor authentication entirely. It started as a personal project. I have written elsewhere about being a bedroom musician who never quite managed to record what I heard in my head. Whether the problem was the tools or the talent is a question I prefer not to examine too closely. overdub was my attempt to build the simplest possible thing for that gap. It runs as a PWA and exports to WAV. It is not trying to compete with professional recording software. It is only trying to stay out of the way.

Unfinished Work

Not every project finds its form.

In 2014 I and a few others began work on a digital humanities network for the region I am from. We called it South Asia Digital Humanities. The idea was to bring together scholars and archivists working across South Asia, along with the technologists who could actually build the infrastructure. It did not work. I was in the right place, just not at the right time. Looking back, it was an early formation, an ancestor to what later became DHAI and then DHARTI.

In 2016 I built a prototype mobile app for exploring historical letters. I presented it at a couple of conferences and there was real interest, but I never released it. Somewhere along the way I lost interest myself. This happens to me often. I write half a paper. I sketch a bit of code. Once the challenge feels solved, I tend to turn away before finishing.

My CV of failures is long. In 2017 I returned to India with a freshly minted PhD and some experience in the European cultural heritage sector. I expected to slot smoothly into the system here. Over the next three four five six years (I gave up counting cleanly after that), I tried to build partnerships with cultural heritage institutions in India. I needed this for my work, and I would argue some of these institutions needed me too. I followed leads and wrote proposals. Nothing came of it. It broke something in me that I do not think I have fully fixed. I have applied to grant bodies and been rejected five times nine times, and counting. Some of that is probably on me. These are the scars I carry. Sometimes I think this is simply what it means to be an academic, and that I am luckier than most.