Work · Vaakh Research

Program 01 · Founding program

Language & literature

Kashmiri is a low-resource language in the technical sense: the corpora, benchmarks, and baselines that language technology is built from largely do not exist for it. Its written tradition spans four scripts across seven centuries; its dominant orthography marks vowels its neighbors do not, which is where systems trained on Urdu and Persian fail; and its printed record sits in encodings no search engine can index. We build the missing foundation: normalization standards, governed corpora, document intelligence for Nastaliq, and speech systems that hear the language while its deepest registers are still spoken.

Program 02

Land

The valley's economy stands in its orchards and fields: apples that supply most of India's crop, saffron that carries Kashmir's name across the world, and diseases and weather that move faster than advice does. We work on computer vision for early detection, on the tree and in the corm, under real field conditions; on forecasting for frost, hail, and yield; and on decision support delivered in the grower's own language, on the phone already in their pocket.

Program 03

Water

The Jhelum has flooded this valley throughout recorded history, and 2014 showed what a week of rain can still do. We work on flood forecasting from rainfall, snowpack, and satellite observation; on monitoring for Dal Lake and the wetlands that once absorbed the river's excess; and on the retreat of the glaciers that feed everything downstream. The unit of success is hours: warning delivered early enough to matter.

Program 04

Health

The valley's health system carries more need than capacity, and distance decides too much. We work on machine intelligence where the system is thinnest: screening support where specialists are few, records and interfaces in the languages patients actually speak, and the quiet infrastructure that lets a clinic in the periphery draw on expertise from anywhere. Everything in this program is built with clinicians, under medical governance, with patient data treated as the most protected class of all.

Program 05

Craft

Authentication is a vision problem, and imitation is an industry. Pashmina and kani carry centuries of technique and a geographical indication the market routinely fakes, at the weaver's expense. We work on systems that can read the hand in the weave: texture, structure, and provenance, established by machine and defensible in trade. The craft economy survives on proof.

Program 06

Memory

Preservation fails at the point of retrieval: what cannot be found might as well be lost. We build archival infrastructure to institutional standard: digitization pipelines, metadata in the valley's own terms, consent frameworks for knowledge held by people rather than institutions, and storage designed for decades rather than product cycles. The goal is a record of Kashmir that outlives every device it was made on.