Dr Jyostna Devi Bodpaati
Development
Andhra Pradesh, India
Skills
Machine Learning (ML)
About
Jyostna Devi Bodapati's skills align with Programmers (Information and Communication Technology). Jyostna also has skills associated with IT R&D Professionals (Information and Communication Technology). Jyostna Devi Bodapati has 14 years of work experience.
View more
Work Experience
ML Expert and Trainer
eWealthMax
December 2022 - Present
- Spearhead the development of sophisticated models tailored for precise tax estimation, leveraging factors like income, deductions, credits, and filing status. • Innovate and implement models adept at discerning tax filing categories and determining eligibility for deductions, enhancing accuracy and efficiency in tax-related services. • Lead training sessions for new team members, imparting expertise in Python, C, and Java programming languages, as well as comprehensive instruction in Data Science, Machine Learning, Deep Learning, and Natural Language Processing (NLP) techniques.Developing models suitable for tax estimation based on factors such as income, deductions, credits, and filing status.
Doctoral Research Scholar
Vignan's Foundation for Science
January 2019 - April 2023
- Vadlamudi, Andhra Pradesh • Attention Based Representation Learning for Limited Data Settings: The primary focus of my doctoral research revolved around the development of attention-based deep learning methodologies, specifically tailored for efficiently learning image representations, especially in scenarios with limited size datasets. • Diabetic Retinopathy Severity Classification: Initially, we devised a feature aggregation framework that integrated CNN-learned features at various levels to create a discriminative representation incorporating low, mid, and high-level features from fundus retinal images. While this approach outperform traditional CNN features, it suffers from limitations such as focusing on lesion-specific regions. To overcome this challenge, subsequent work introduces a composite deep neural network with gated attention, excelling in capturing lesion-specific features. However, handling higher-dimensional attention features poses a challenge. In the subsequent studies, we introduced a joint neural network comprising a convolutional auto-encoder and a discriminator to compress these high-dimensional attention features. Additionally, we designed a Neural Support Vector Machine as a hybrid model combining the strengths of neural networks for representation learning and SVM for robust classification. These efforts collectively led to a significant 9% improvement in diagnostic accuracy for retinopathy severity grading compared to existing research. • Brain Tumor Recognition: Our initial approach involves a joint model capable of learning distinct features from brain MR images, facilitating the identification of brain tumor types. The key strength of this model lies in its ability to assimilate diverse features from two different channels. An extension of this work introduces Multi-modal Squeeze and Excitation based attention, enhancing the focus on tumor-specific regions across the features. Seeking further performance enhancement, subsequent work results in a multi-level attention network featuring various types of visual attentions. This design aims to capture spatial and channel-specific attention scores in relation to tumored regions, resulting in a notable 6% improvement in precision.
Assistant Professor
Vignan's Foundation for Science
June 2016 - December 2019
- (deemed to be) University ∗ Rich expertise in teaching cutting edge technology courses such as Machine Learning, Deep Learning, NaturalVadlamudi, Andhra Pradesh Language Processing and Pattern Recognition ∗ Trained students on Computer Science foundation courses such as Data Structures & Algorithms, Object Oriented Programming in C++ & Java, Software Engineering Methodologies and Web Technologies ∗ Mentored UG & PG Students in developing Major Projects by following the latest IT trends ∗ Trained faculty on high-demand courses such as Python Programming and Machine learning ∗ Committed to staying up-to-date with the latest developments in the field, completed online certifications offered by premier institutes, proactively participated in several faculty development programs, workshops ∗ Developed video lectures for on-demand courses such as Python Programming, machine learning and Pattern Recognition ∗ Demonstrated strong organizational and communication skills by organizing workshops and faculty development programs, coordinated university level technical events
Research Assistant
Indian Institute of Technology
June 2013 - May 2016
- Developed semi-supervised approaches to deal with limited label data problems that are suitable for image andChennai, Tamil Nadu speech data ∗ Served as teaching assistant for courses such as Data structures and Algorithms, Pattern Recognition, Kernel Methods for Pattern Analysis ∗ Published research outcomes in International journals of high repute
Assistant Professor
Vignan's Engineering College
June 2007 - November 2007
- ∗ Taught Object Oriented Programming through C++ and Web Technologies Subject to UG Students Vadlamudi, Andhra Pradesh ∗ Attended Personal Development & Soft-skills for teaching programs
Lecturer
Vignan Degree & PG College
August 2005 - June 2007
- Delivered courses such as C Programming, Unix & Shell Programming and Software Engineering subjects to UGGuntur, Andhra Pradesh and PG Students
Software Engineer Trainee
ProdEx Technologies
December 2004 - August 2005
- Developed C++ Programming logic according to Functional requirements
Education
Technology & Research University
Doctor of Philosophy in Computer Science & Engineering
JNTU
Master of Technology in Computer Science & Engineering
