Browse some of my featured projects. Click any card to open the full project page, read the details, and view images or links. I have only highlighted some of my recent work here, you can find more details on my older work in my CV, which is also attached.
We applied causal inference methods such as difference-in-differences and triple-difference models to evaluate the impact of COVID-19 on child nutrition outcomes across caste groups using India’s NFHS-5 survey data.
Working with the World Bank Health Group, we developed a geospatial framework integrating clinic locations, road networks, and internet connectivity data to assess healthcare accessibility across South Africa.
We’ve built the first open-source African stereotype dataset and evaluation framework grounded in local socio-cultural contexts, to evaluate LLMs for Afro-specific bias.
We have come up with a methodology through which Automatic Speech Recognition (ASR) models can be fine-tuned for specific domains for low-resource languages such as Wolof and Hausa. Specifically, we have fine-tuned these models for the maternal and reproductive healthcare domains.
In this project, we put forward a methodology through which we can study caste bias in state-of-the-art multilingual LLMs, examining how models associate caste-indicative names with positive and negative attributes across multiple Indian languages.
The ‘Red Corridor’ in India is considered a media blackout zone, with there being very little media reporting from the area. We attempt to use remote sensing techniques to paint a picture of the constant confict that is happening in these regions.
