Hi, I'm Sarat

Introduction

Overview.

Hello! I’m a Computer Science graduate student pursuing an M.Tech in Computer Science, with interests in systems, computer architecture, and performance-oriented computing. I have hands-on experience with dynamic binary instrumentation tools such as DynamoRIO and Intel PIN, and enjoy exploring low-level systems to understand and improve real-world software behavior. Currently, I’m learning about LLM inference performance, including basic profiling and performance analysis across CPU/GPU systems, and I bring strong problem-solving skills, adaptability, and a collaborative mindset to my work.
Let's connect and build something impactful together!

gandholisarat@gmail.com

+91 77308 34575

Resume

DBI Instrumentation

Problem Solving

Pure Maths

Leadership

What I have learnt so far

Skills.

My work

Projects.

These projects showcase my practical skills and experience, each with descriptions and links to code repositories and live demos. They demonstrate my ability to handle complex challenges, adapt to different technologies, and oversee projects from start to finish.

DynamoRIO Custom Client

A custom dynamic binary instrumentation (DBI) client built using the DynamoRIO framework. This tool is designed to analyze and modify program behavior at runtime for performance monitoring, security research, or reverse engineering tasks.

#C/C++

#DynamoRIO

#binary instrumentation

#systems programming

LLM from Scratch (Learning Project)

A learning-focused project to understand large language models by implementing core components from scratch, including tokenization, embeddings, attention mechanisms, and basic transformer architecture, with an emphasis on inference behavior and performance considerations.

#Python

#Machine Learning

#Transformers

#LLM fundamentals

#learning project

Feedforward Neural Network

An educational Python implementation of a feedforward neural network from scratch, designed to deepen understanding of forward and backward propagation, weight updates, and loss calculation without using deep learning libraries.

#python

#neural networks

#machine learning

#backpropagation

#from scratch

Java Server-Client

A simple Java-based server-client architecture demonstrating socket programming. This project enables multiple clients to connect to a server and exchange messages, making it a great starting point for understanding networking in Java.

#java

#socket programming

#networking

#multithreading

#tcp/ip

My Writings

Blog Posts

Get in touch

Contact.

I'm open to internships, collaborations, and exciting projects. Let's connect and build something meaningful together. Feel free to check out my resume and reach out!