Our research programme explores the unique data landscapes, language challenges, and socioeconomic contexts that shape effective AI for Asian populations.
Developing NLP and language model techniques that work effectively for Bangla, and other South/Southeast Asian languages with limited training data.
Computer vision and predictive modelling techniques adapted to the crop varieties, disease patterns, and climate conditions of South Asia.
Model compression, edge deployment, and offline-first architectures that deliver AI capability without reliable high-speed internet.
Analysing the structure of Asian job markets, skill demand signals, and career trajectory patterns to power career intelligence products.
Understanding how people with varying levels of digital literacy interact with AI interfaces, and designing for genuine inclusion.
We publish research notes, findings, and working papers as we build. All work is open — we believe AI knowledge for Asia should be shared.
An overview of existing Bangla NLP datasets, model performance benchmarks, and the specific challenges of building production AI systems for the Bangla language in 2025.
Request access ↗Analysis of the structural mismatch between graduate skill profiles and employer expectations in Bangladesh, with implications for AI-powered career guidance tools.
Request access ↗Results from testing MobileNet-based crop disease classification models on sub-$50 Android handsets typical of rural Bangladesh. Accuracy, latency, and offline performance benchmarks.
Request access ↗An argument for why the most societally significant AI applications of the 2020s will emerge from South and Southeast Asia — and what it will take to build them.
Read on the blog ↗We are open to collaborating with universities, research institutions, and domain experts across Asia on problems that matter.