Projects
This section collects things I've built or am actively building: tools, prototypes, technical experiments, and applied systems work. Unlike Writing or Notes, the center of gravity here is making rather than explaining, and it is most useful for readers who want to see how ideas become working artifacts.
A WhatsApp-first AI assistant for Indian farmers — answering questions on fertilizers, mandi prices, weather, and crop health in their own language, where they already are.
Problem
India's 150 million smallholder farmers lack timely, personalized agronomic advice in their own language. Extension services are thinly spread, most agri-apps assume smartphone literacy and English, and mandi price information passes through layers of middlemen.
Outcome
A WhatsApp-native AI assistant that speaks the farmer's language, gives real-time answers on crops, inputs, market prices, and weather — meeting farmers on the platform they already use every day.
An AI-powered Python learning web app with an in-browser code editor and a conversational AI tutor that generates lessons and explanations on-demand — built so beginners can learn by doing and asking, not by watching.
Problem
Standard Python courses are passive and non-adaptive — video lectures and static exercises leave beginners stuck without anyone to ask. A motivated learner with a specific question has to wade through documentation or wait for a forum reply.
Outcome
A web app where you write Python in the browser, ask the AI tutor anything mid-session, and get a direct explanation or a mini-lesson on-demand — turning learning into a live conversation rather than a fixed curriculum.
Building biochar-based biological agri-input products for Indian farms — adding stable carbon to soil, restoring soil biology, and improving climate resilience for smallholder farmers.
Problem
Indian smallholder farms face declining soil organic matter, rising input costs, and increasing climate stress — while conventional agri-inputs do little to restore the soil biology that underpins long-term land productivity.
Outcome
Developing advanced biochar-based biological agri-inputs that add stable carbon to soil, support beneficial microbiology, and improve drought resilience — being trialed on family farms in collaboration with a cousin.
An interactive visual lab embedded in the Newton–GMRES article — explore solver traces from a 60-species stiff chemistry system and see why preconditioning collapses GMRES iteration counts from 172 to 4.
Problem
Solver internals are opaque from text alone. The interaction between nonlinear residual reduction and inner linear convergence — and the dramatic impact of preconditioning — is best understood by seeing real solver traces directly.
Outcome
An interactive lab embedded in the full article, giving readers a hands-on way to explore Newton outer iterations, GMRES convergence curves, and the sparsity structure that explains why banded preconditioning wins.