hey, i'm dhruva. i'm 19, currently in my first year of undergrad.

i'm a quantitative researcher at Basalt Research and Technologies, where i design and evaluate forecasting models for NIFTY-50 and India VIX — trying to find signals that actually hold up under regime shifts, not just in backtests. most signals don't survive. that's what makes the ones that do interesting.

i also do ML research. my recent work is PhysicsDiff-SSM — a physics-constrained latent diffusion model for industrial bearing fault signal generation. i introduced PVR% (Physics Violation Rate) as a metric for physical consistency of generated samples.

right now i'm building NIL — software that makes NHCX (India's national health claims exchange) actually usable for hospitals. most providers are still managing insurance claims through Excel and WhatsApp. NIL centralises the insurance desk workflow and validates submissions before they go out. the longer play is building data intelligence on why claims fail. we were recently covered in First Post.

mathematically, i was selected to the Indian Statistical Institute mathematics camp (ranked 27th nationally, category B, among 1000+ participants). i'm currently working on an independent preprint on sign analysis of generalised Fibonacci sequences — specifically, closed-form conditions on alternation in linear recurrences and the geometry of alternation strips in parameter space.

i got into all of this early. i'm interested in things that are hard to fake: mathematical structure, signal that survives out-of-sample, models that respect physics. i tend to distrust results that are too clean.