LLM Cache Engineering
An open-source caching layer for Large Language Models that remembers and reuses previous responses to similar queries. Built to reduce costs and latency when working with LLMs at scale.
Projects
Adaptive Randomized Smoothing: Certifying Multi-Step Defences against Adversarial Examples Research
We propose Adaptive Randomized Smoothing (ARS) to certify the predictions of test-time models against adversarial examples. ARS extends the analysis of randomized smoothing using f-Differential Privacy to certify the adaptive composition of multiple steps.
ML Privacy Meter Engineering
ML Privacy Meter is a Python library that enables quantifying the privacy risks of machine learning models. The tool provides privacy risk scores which help in identifying data records among the training data that are under high risk of being leaked through the model parameters or predictions.
GMRT Archive Processing Project Engineering
We have built a high performance compute cluster and data science pipeline to synthesize images from raw interferometric data collected by the Giant Metrewave Radio Telescope (GMRT). Our efforts helped reduced synthesis time from 5 months to around 1 month. Project is currently generating one of the world's largest catalogs of sub-GHz frequency radio astronomy images.
Global Approximation of linear and polynomial functions on Manifolds via Partitions of Unity Research
In this work, we empirically test a hypothesis that locally fitted linear and polynomial functions on the charts of a 2-dimensional manifold can be globally approximated over the whole surface of the manifold. For doing so, we consider the Partitions of Unity method.