Satvik

Research Commons

AI Research Intern • Jan, 2025 — Present

Working on Agentic AI for autonomous research agents;

College Setu

ML Enginner Intern • May, 2023 — July, 2023

Worked on creating a chatbot for the College Setu website and also worked on scraping data for training the model.

BITS Pilani, Goa

B.Tech. in Electronics and Instrumentation • 2021 — Present

Currently doing multiple machine learning projects and research work at BITS Pilani, Goa.

St. Michael's High School, Patna

High School • 2020

Scored 92.8% in CBSE Board Exams.

LLMs for real time network traffic analysis

Researcher • 2023 — Present

We have created a foundational model compatible with network data. Model is trained on a huge hex-dump dataset on A100 GPUs. Currently, we are working on fine-tuning it on different datasets for bencharmarking.

Undergraduate Thesis

• August, 2024 — Dec, 2024

I completed my undergraduate thesis on the topic of metabolomics research with machine learning. The the usage of novel deep learning algorithms wasn't possible due to the size of the dataset. While working on the analysis of GC-MS data I faced many difficulties in learning complex softwares. So (on recommendation from my advisor) I created a library called 'mbSTATS' which is a collection of all the tools and techniques required for the pre-liminary analysis of GC-MS data. The library is written in Python and is open-source. The library is currently being used by the metabolomics research group at BITS Pilani.

Kaggle Contributions

• 2023 — Present

Optimized the inference time of Mistral-7B model to achieve a throughput of 300 tokens/sec on a RTX 3050ti GPU with the help of model pruning and quantization.

Learning Agency Lab - Automated Essay Scoring

• 2023;

Fine-tuned roberta to acheive a public score of 0.777 where the top scorer had a score of 0.82.

Electronics store management (OOPS project)

• 2023;

This involves managing an Electronics Store model, offering a diverse range of electronic products such as air conditioners, refrigerators, televisions, washing machines, and dishwashers. I am tasked with designing and developing a comprehensive system to handle inventory management, including tracking incoming and outgoing products and updating product details with purchase and sales information.

Part I: Introduction to MoE (Mixture of Experts)

2024-01-19 •

First article in the exploration of journey of MoE's from scratch to GPT-4. In Part I: we have explored a basic structure of MoE and their implementation in pure numpy.

Timeline of Computer Vision

2025-01-04 •

Survey and Comparison of models which can be used for Computer Vision tasks. In this article, I have compared three models: a Simple CNN model, VGG16, and ViT for an image classification task.

Thesis Report

2024-12-12 •

My thesis report explaining the usage of machine learning algorithms in metabolomics and the creation of the data analysis library 'mbSTATS'.

GSOC'24 proposal (Rejected)

2024-12-12 •

My proposal for GSOC'24 that got rejected. (A lesson for someone applying - don't apply in institutions which have more than 20 projects because only 7 or 8 projects get selected from each institution.)

NetLM - A foundational model for network data

• 2024

Creating SOTA foundational model based on hierarchical transformers for network data.

Programming languages

Python, Java, SQL, MATLAB

Frameworks and libraries

Pytorch, tinygrad, Scikit-learn, Pandas, Numpy, Matplotlib, Huggingface Transformers, Keras

Deep Learning

LLMS, CNNs, RNNs, Transformers, GNNs

• Painting and Sketching