Hello! I'm

BHAVYA MEHTA

Bachelors in Computer Engineering

Software Developer | ML Enthusiast | Learner and Researcher

Download CV

ABOUT

Let me Introduce Myself

I am Bhavya Mehta, an aspiring Computer Engineer passionate about building impactful technology.

I hold a Computer Science degree from VJTI, where I built strong fundamentals in Systems, Algorithms, and Software design.
I bring 2 years of full-time experience in Full Stack Development, having worked at JP Morgan & Chase and Kelp Global, where I developed scalable systems and contributed to key product features.

My research experience at Stanford University, IIT Delhi, IIT Kharagpur includes work on ML in Drug Discovery, Quantum Materials, and Vision Transformers, expanding my expertise in data-driven innovation and ML Research with 4 publications in top international conferences like FedCSIS, IEEE IRI and more.

I would love to connect if you seek a passionate problem-solver with a knack for innovation and a drive to excel.
Let's explore the possibilities together and make a meaningful impact in the world of technology!

Dive down to read more about me!

EDUCATION

Stony Brook University

2025-2027 (Expected)

Master of Science in Computer Science

Algorithms for Data Science, Distributed Systems, Advanced NLP, Information Retrieval, ML.

Veermata Jijabai Technological Institute

2019-2023

Bachelor of Technology in Computer Engineering

CGPA: 9.42 ( Department Rank 3)
Applied Mathematics, Operating Systems, Data Structures, Algorithms, Linear Algebra, Databases, ML.
Final Project: Toxic Molecule Classification Using Graph Neural Networks and Few Shot Learning.

EXPERIENCE

Check out my Work & Research Experiences

Kelp Global

Oct 2023 - Present

Software Engineer

Recognized as Best Performer (Q3 2023) for leading product development and elevating client satisfaction through impactful full-stack contributions.
Delivered 30+ end-to-end features, optimized Angular builds with Micro Frontends (75% faster), and improved UX by reducing component load times (52%) via ECharts integration.
Engineered scalable features like Export to Excel & Peer Comparison for 80K+ rows, and migrated 4TB SQL data to Parquet on AWS, saving $10K annually.
Automated dependency management by building a custom GitHub Dependabot, saving 10+ manual hours monthly and enhancing CI/CD workflows.
Led as both Product Manager and Sprint Master, defining user stories, engaging with clients, and driving a 10-member team through agile cycles and production releases.

VJTI

Jan 2023 - Sept 2023

Research Assistant

Collaborated with Prof. Udmale to develop an ML-based discovery engine for predicting Heusler compounds.Novel approach is 50x faster than diffraction methods & predicts 144 new Heuslers with a 0.99 TPR.
Modelled regression based analysis on the the co-variation of DNA to RNA, and to surface proteins in single cells during hematopoietic stem cell development, under the mentorship of Prof. Chandane.
Trained Echo State Networks, on Multiomic and CiteSeq datasets resulting in a SOTA correlation score of 0.94 and 0.895 enhancing processes in drug discovery & disease biomarker detection.

Stanford University

Nov 2022 - Mar 2023

Research Collaboration

Conducted an in-depth literature review on Invertible Neural Networks (INNs), synthesizing applications across cryptography, data compression, and other domains.
Developed a dynamic knowledge graph timeline to visualize and track the evolution of INN research methodologies.

IIT Delhi

Oct 2022 - Mar 2023

Research Assistant

Explored Conditional Prompt Learning for Vision-Language Models, transforming context tokens into learnable vectors; surveyed advanced methods like CoOp, CoCoOp, and TPT.
Achieved 97% accuracy on datasets like StanfordCars & Flowers102 by integrating a Multi-Headed Self-Attention model, improving feedback aggregation and boosting learning rate by 2%

Fellowship.AI

October 2022 - December 2022

Research Fellow

Presented solutions to backdoor attacks on Deep Neural Networks, including defenses against the COLLIDER method.
Contributed to an iOS app investigating links between reduced Heart Rate Variability and food sensitivity, supporting data collection and analysis.

IIT Kharagpur

July 2022 - Dec 2022

Research Assistant

Pre-processed clinical data for Laparoscopic Cholecystectomy prediction using Knowledge Graphs, optimizing node and edge coordinate assignments with a recursive child rank algorithm.
Reduced visual overlaps by 80% and cut per-node processing time from 0.2ms to 0.09ms, significantly enhancing system efficiency and clarity.

JP Morgan & Chase

May 2022 - July 2022

Software Engineer Intern

Developed a Certificate Management Tool to recursively validate keystore files, automate expiry monitoring, and deliver real-time UI/email alerts—reducing manual tracking by 80%.
Optimized keystore handling by caching paths and passwords, cutting server storage by 50% and improving system reliability with 20% fewer production disruptions.

SKILLS

I can do well in...

● Programming Languages

C++

Python

Java

GoLang

Typescript

● Frontend Frameworks

HTML

CSS

SASS

Javascript

Bootstrap

JQuery

React.js

Angular

Flutter

Dart

● Backend Frameworks

Node.js

Express.js

Nest.js

Spring

Flask

● Databases

SQL

PostgreSQL

MongoDB

ElasticSearch

Firebase

● Data & ML Frameworks

SkLearn

Keras

PyTorch

Hadoop

NetworkX

● Tools

Git

Docker

Jenkins

PUBLICATIONS

Check out my research work

19th Conference on Computer Science and Intelligence Systems FedCSIS

Core Rank B | 2024

Toxic Molecule Classification Using Graph Neural Networks and Few Shot Learning.

Traditional methods like Graph Convolutional Networks (GCNs) face challenges with limited data and class imbalance, leading to suboptimal performance in graph classification tasks during toxicity prediction of molecules as a whole. To address these issues, we harness the power of Graph Isomorphic Networks, Multi Headed Attention and Free Large-scale Adversarial Augmentation separately on Graphs for precisely capturing the structural data of molecules and their toxicological properties. Additionally, we incorporate Few-Shot Learning to improve the model's generalization with limited annotated samples. Extensive experiments on a diverse toxicology dataset demonstrate that our method achieves an impressive state-of-art AUC-ROC value of 0.816, surpassing the baseline GCN model by 11.4\%. This highlights the significance of our proposed methodology and Few Shot Learning in advancing Toxic Molecular Classification, with the potential to enhance drug discovery and environmental risk assessment processes
DOI: http://dx.doi.org/10.15439/2024F3810

IEEE 24th International Conference on Information Reuse and Integration for Data Science

Core Rank C | 2023

Accelerating the Search for Stable Full Heusler Compounds through Machine Learning

Applications for Heusler compounds are expanding in topological insulators, magnetocaloric, spintronics, and superconductivity areas. These substances are expanding the boundaries of science and offering answers to engineering problems. Our work demonstrates a discovery engine that can predict the crystal structures and chemical characteristics of 1107 Full Heusler compounds by implementing a Machine Learning approach trained with elemental descriptor data. Our approach is 50 times faster than rule-based and diffraction techniques, with a true positive rate of 0.99 for every random combination of elements on more than 1,000,000 candidates. We also compute the formation energies of these novel compounds to filter out 144 highly stable Heuslers that coincide with existing research and density functional theory trends to validate and support our findings.
DOI: https://doi.org/10.1109/IRI58017.2023.00034

14th International Conference on Computing Communication and Networking Technologies (ICCCNT)

2023

Exploring Graph Classification Techniques Under Low Data Constraints: A Comprehensive Study

This survey paper presents a brief overview of recent research on graph data augmentation and few-shot learning. It covers various techniques for graph data augmentation, including node and edge perturbation, graph coarsening, and graph generation, as well as the latest developments in few-shot learning, such as meta-learning and model-agnostic meta-learning. The paper explores these areas in depth and delves into further sub classifications. Rule based approaches and learning based approaches are surveyed under graph augmentation techniques. Few-Shot Learning on graphs is also studied in terms of metric-learning techniques and optimization-based techniques. In all, this paper provides an extensive array of techniques that can be employed in solving graph processing problems faced in low-data scenarios.
DOI: https://doi.org/10.1109/ICCCNT56998.2023.10307388

International Conference on Intelligent Computing, Simulation and Optimization

2023

Regression-Based Analysis of Multimodal Single-Cell Data Integration Strategies

Multimodal single-cell technologies enable the simultaneous collection of diverse data types from individual cells, enhancing our understanding of cellular states. However, the integration of these datatypes and modeling the interrelationships between modalities presents substantial computational and analytical challenges in disease biomarker detection and drug discovery. Established practices rely on isolated methodologies to investigate individual molecular aspects separately, often resulting in inaccurate analyses. To address these obstacles, distinct Machine Learning Techniques are leveraged, each of its own kind to model the co-variation of DNA to RNA, and finally to surface proteins in single cells during hematopoietic stem cell development, which simplifies understanding of underlying cellular mechanisms and immune responses. Experiments conducted on a curated subset of a 300,000-cell time course dataset, highlights the exceptional performance of Echo State Networks, boasting a remarkable state-of-the-art correlation score of 0.94 and 0.895 on Multi-omic and CiteSeq datasets. Beyond the confines of this study, these findings hold promise for advancing comprehension of cellular differentiation and function, leveraging the potential of Machine Learning.
DOI: https://doi.org/10.48550/arXiv.2311.12711

PROJECTS

Check out my work

Fake Face Generator

ML DL |

Implemented a research based DCGAN model to generate fake human faces addressing the problem of image data scarcity. Carried out distributed training of 10,000+ images with an average discriminator-generator loss of 0.76 and 0.89. The model has a total of 29 million parameters & outputs unique realistic faces after 30 epochs of training. An ensenmble of generators pair up with a discriminator to extract facial features and distinguish between real and fake images.
Technology Stack: Python, Sklearn, Keras, Pytorch, GPUs.

Dyce & Dyne

AI Algos |

A modern day food ordering website which implements algorithms like A*, Minimax and alpha-beta pruning for 3 game based wallets instead of Promo codes. Solved the TSP using Ordered Crossovers with mutation at a rate of 0.05% and used Mapbox APIs to show the optimal route for delivery considering real time traffic.
Technology Stack: HTML, CSS, Js, MongoDB, Express.js, Node.js

AID

Charity |

A complete, end‐to‐end solution for the functioning and manage‐ ment of an NGO. Includes Donate, Volunteer, Ecommerce, Blogging, E‐Payment and Mailing features.
Technology Stack: HTML, CSS, Js, MongoDB, Express.js, Node.js

HisabKitab

FinTech |

HisabKitab is a simple GST Billing and Stock Management software.It is a systematic user centered platform for local vendors to create,store manage and download structured invoices as pdf for all transactions fed. Provides multi-chart statistical analysis for monthly sales and purchases along with fuzzy search to easily navigate through documents. Helps keeping a track of transactions and printing bills at the same time.
Technology Stack: HTML ,CSS ,Js, Express.js, Node.js,Google Firebase, Dart, Flutter

Purrfect

Social Media |

An all inclusive community and shopping site for animal lovers of all stripes and colors. Users can create Facebook‐style profiles for their pets and interact with other users.
Technology stack: HTML, CSS, Js, MongoDB, Express.js , Node.js

Sahara

Social Cause |

A Food donation app that connects potential food doners to NGOs & receivers preventing wastage of food and starvation. Built end-to-end workflow including use cases such as user sign up, profile creation,donor-donee matching and conditional data retrieval for filtering. Conducted usability testing on 15+ users to enhance user experience. A step towards eliminating starvation and food wastage.
Technology stack : Dart, Flutter, Google Firebase

Jarvis

Automation |

Voice activated assistant which takes commands from the user and works accordingly. Does day to day tasks for user like writing notes, sending emails,sending messages over whatsapp. And for it to interact with the user ,it uses Speech recognition ie the process of converting audio into text.
Technology Stack: Python, SpeechRecognition.

I would love to hear from you. Feel free to reach out.

Mumbai, Maharashtra

bhavyamehta922@gmail.com