I'm a Computer Science Master's student at Columbia University, on the Machine
Learning track, with a 4.0 GPA and a PDL Fellowship. My work lives at the point
where applied ML meets the constraints that keep real systems honest -
encrypted data that can't be decrypted to be searched, weather stations that
disagree with each other, benchmarks that quietly undercount their own
failures.
Most recently at Protegrity, I built a privacy-preserving
knowledge graph search system using homomorphic encryption that never exposes
plaintext, even to the model doing the reasoning. At Columbia's
WiMNet Lab, I engineered a spatial interpolation pipeline
turning a noisy mesh of personal weather stations into 15-minute rainfall maps
for NYC. Before Columbia, I spent over a year as a software engineer at
Syngenta shipping a production React Native app, and earlier
built explainable AI tools for oncology drug recommendations and crop disease
detection.
Here are a few technologies I've been working with recently:
- Python
- LLM Agents
- AI & ML
- React / React Native
- SQL & BigQuery
- Homomorphic Encryption
const sanskruti = {
role: 'MS CS - ML @ Columbia',
focus: [
'generative ai',
'machine learning models',
'explainable AI'
],
gpa: 4.0,
publications: 2,
recognition: 'CancerXAI copyright grant',
openTo: 'full-time, Dec 2026',
};