Worky | Meet our Talent - Abhishek Kumar

Skills

Data Science

About

Abhishek Kumar's skills align with Programmers (Information and Communication Technology). Abhishek also has skills associated with IT R&D Professionals (Information and Communication Technology). Abhishek Kumar appears to be a low-to-mid level candidate, with 3 years of experience.

Work Experience

Software Engineer

Carpl.ai

December 2021 - July 2022

Work on building the dashboard for x ray and diacom images. Fixed the bugs and developed the API for the CARPL dashboard using Django, nginx, docker and react js. Maintain and reduced the api call time using Fast api in some cases.

Associate Data Engineer

Innoplexus Consulting S

August 2021 - December 2022

ervices Focused on data crawling and data extraction from different types of htmls and pdfs. Port and clean the crawled data into the CLS of different asset classes like:-Publications(PubMed), Congresses and News. Using Elasticsearch query fetch the data from different asset classes for product development and analytics. Extract the unstructured pdfs text data from different sources using a pdf to text conversion and after this string manipulation for all sources. Worked on hospital wise crawling of physicians data from different sources in the Curia App development. Tools:-Beautifulsoup, Selenium, Elasticsearch, Mongodb and BERT.

Senior Data Scientist

BAJAJ FINSERVE

August 2022 - Present

Work on multiple projects on AI and NLP. Focused on pdf extraction and cleaning of various bank statements having different formats on each bank. Work on OCR project where build deep learning model for bank logo detection using YOLO V7 model, accuracy of the model is around 98 percent. Work on bank demographics extraction from pdf and images inside bank statement. Build dashboard for Perfios banking index using Django and React js.

Machine Learning Engineer(Internship) at Quantiphi

Analytics Solutions

May 2019 - July 2019

Focused on research project to reduce the training time of machine learning classification model(ResNet50) on image datasets using clusters of multiple GPUs. Using Distributed Deep Learning with Horovod technique after benchmarking of ML mode ltraining time reduced by two third. Benchmarking and accuracy measured by F1 score and confusion matrix. Tools &Technologies:-Google Cloud Platform(GCP), Keras, Tensorflow, Python, Ring all reduce algorithm, Nvidia Tesla k- 80, GPUs, Horovod, Supervised Learning, Distributed Deep Learning, ResNet- 50. Worked on OCR and BERT sentence transformer model for detect the text from the TV data images and convert it into english language.

Undergraduate Project at Indian Institute of Technol

January 2019 - March 2019

based hotel reviews using supervised learning methods. Using NLTK and ML trained a model which gives the positive or negative sentiment forthe given reviews and ratings. It can rate the Hotels, PG, Resturants etc by using customer reviews. Used classifier algorithms are:- NaiveBayes, SVM, DecisionTree and Logistic regression. Further accuracy increased by Ensemble Learning. Tools:-Python, NLTK, Confusion Matrix, Sklearn.

Abhishek Kumar