Karthik Ravichandran

PeerJ Reviewer

Summary

⚡ Karthik Ravichandran, a dedicated professional hailing from India, currently pursues his graduate studies at the prestigious University of Massachusetts, MA, USA. As a Graduate student, he is currently focusing on the personalization of Language Model Libraries (LLM) using Information Retrieval. This research, initially undertaken in CS646 under the guidance of Professor Hamad Zamani, reflects Karthik's commitment to advancing the field.

⚡ With a rich background spanning five years in the dynamic field of Data Science, Karthik has significantly contributed to Image and Signal Processing, Bio-medical research, and Natural Language Processing.

⚡ Commencing his career with a noble cause in mind, Karthik's initial foray into the professional world led him to the Healthcare industry following his Bachelor's degree. Focused on solving Cardiovascular problems, particularly the prediction of Ischemic diseases, he applied state-of-the-art techniques, ultimately saving approximately 5000 lives per month through critical alerts derived from ECG analysis. Beyond algorithm development, Karthik actively engaged with cardiologists to enhance his understanding of ECG diagnosis.

⚡ Continuing his trajectory within the Healthcare sector as a Data Scientist, Karthik transitioned his expertise from core signal processing to Natural Language Processing (NLP). Undertaking the challenge of predicting ICD-10 codes from medical charts—comprising a vast array of classes exceeding 10,000—Karthik navigated through a skewed dataset. In the absence of open-source Language Model Libraries (LLM), he developed solutions using attention-based architectures like CAML, leveraging in-house GPU machines. Overcoming challenges posed by chart length, he extended transformer architectures and explored techniques such as Transformer XL. To validate model predictions, he implemented a combination of classification and term search algorithms, including sentence/word vectorizers.

⚡ Subsequently, Karthik seized an exciting opportunity with Walmart Global Tech, a Fortune 1 company and a tech giant. As a Data Scientist, he played a pivotal role in not only developing algorithms but also establishing connections with internal business partners and directors. His responsibilities included working on low-level algorithms, guiding and training interns and junior associates, and contributing to the positive impact of the company. Recognized for his exceptional contributions, Karthik received a Bravo award for filing Intellectual Properties (IPs) related to clustering algorithms used for recommendations. Furthermore, he created internal tools and Python packages widely adopted by Data Scientists across Walmart.

Artificial Intelligence Computational Linguistics Computer Vision Data Mining & Machine Learning Data Science Natural Language & Speech Neural Networks Sentiment Analysis

Past or current institution affiliations

University of Massachusetts at Amherst

Work details

Data Scientist

Walmart Global Technology

ended - May 2026

Applied AI

Data Scientist

Tricog Health

November 2018 - May 2020

AI Algo Development

• Development of AI Algorithm in C++ and Python to classify and identify cardiovascular criticalities using 12-lead ECG data • Improved the T-wave detection using a feedback mechanism, and the detection of Non-specific T-wave abnormality. • Developed a classifier for T-wave-based diseases using a combination of LSTM and CNN2D • Enhanced beat classification model using unannotated data by implementing Semi-supervised learning and active learning pipelines • Fine-tuning AI models to detect Ischemic conditions like LT, AT, IT, and other T-wave dependent conditions like supraventricular tachycardia in ECG records • Event and Activity analysis of 25 doctors from Tricog hub using a viewer tracking system which includes: a. Building a recommendation engine for diverting ECGs to respective doctors, b. Central tendency analysis of diagnosis time for normal, abnormal, and critical ECGs with respect to doctors

Software Engineer

Omega Healthcare

August 2021 - December 2021

Data Science and NLP

• Built a custom deep learning architecture to classify ICD-10 codes in medical charts (10k classes) • Used Python, PyTorch, and other data science libraries to train, test, and productize NLP models. • Created a method that helps the system infer patterns from old data and use newer data to train a DL model- (self-supervised learning) • Modified Fasttext, CAML, BioNER, and SGM for Multi-label Classification and entity recognition. • Created robust Data pipeline to flow data from DB to OCR, parsing and prediction engine in OSCAR product

Data Scientist

Walmart Global Tech (Walmart Labs)

December 2021 - August 2023

NLP and Data Ventures

• Supported the development of an NLP suite that generated Walmart around 11M USD last FY. • Developed an AI framework (with UI and sophisticated data flow) that makes non-technical business partners use various trained NLP models and other data solutions. • Developed a framework (MR-TOPIC) for topic analysis that’s not only unique but helps Walmart’s business heavily - it’s been considered for filing patents. • Key contributor to customer experience NLP research that focuses on Meta-Learning and few-shot mechanism - Mainly working on Table to Text, Summarization, and Question-Answering • Developed Seller Centric Classification model for Marketplace-Seller feedback analysis. • Mentored 4 Interns and 1 Full-time Walmart Associate in the Data Science team • Productionized 7 international business workflows and developed a robust fallback approach using Airflow

Graduate Student

University of Massachusetts at Amherst

September 2023 - July 2025

Computer science

Websites

PeerJ Contributions

Reviewed 1

Signed reviews submitted for articles published in PeerJ Note that some articles may not have the review itself made public unless authors have made them open as well.

February 28, 2024

A neural machine translation method based on split graph convolutional self-attention encoding

Fei Wan, Ping Li

https://doi.org/10.7717/peerj-cs.1886