Parmanand Sahu

Parmanand Sahu

Computer Science Grad

The University of Texas at Dallas

Biography

As a Principal Data Scientist at Capital One, with over 8 years of experience in the field of data science, I am utilizing my expertise in deep learning and machine learning to predict mortgage prepayment rates for liquidity and interest rate risk management. My background includes extensive experience in data analysis, feature engineering, and optimization techniques, as well as development of internal packages to manage the workflow of large-scale experiments. My work has resulted in significant savings for the company as it is used to project cashflow, which is crucial for risk measurement and valuation of Capital One’s massive MBS portfolio (worth over 70 billion dollars) and new purchases.

I have also worked as a Data Scientist Intern at Ocrolus, where I debugged and verified a data engineering pipeline for document parsing, and as a Data Scientist and Platform Development Intern at Lantern Pharma where I developed a data pipeline for Cancer Drug exploration platform.

I also have experience as a Data Science Intern at Capital One, Machine Learning Specialist at VHSS Lab, University of Texas at Dallas, Artificial Intelligence Engineer at Huddl.ai and as an Associate Data Science Engineer at CoArtha Technosolution. I am proficient in various technologies and tools such as Python, Optuna, AWS, Pytorch, Node.js, and Elastic-Search.

Throughout my career, I have worked on a variety of projects utilizing machine learning and deep learning techniques and have been able to deliver tangible results for the companies I have worked for. I am excited to continue to use my skills and experience to drive innovation and drive business growth.

Interests

  • Deep Learning
  • Natural Language Processing
  • Computer Vision
  • Risk Modeling
  • Digital Marketing

Education

  • MS in Computer Science, 2021

    The University of Texas at Dallas

  • B.Tech in Metallurgical Engineering, 2013

    National Institute of Technology,Raipur

Experience

 
 
 
 
 

Principal Associate, Data Scientist

Capital One

Jan 2023 – Present Mclean,VA

Capital One Financial Corporation is an American bank holding company specializing in credit cards, auto loans, banking, and savings accounts operated primarily in US.

Responsibilities:-

  • MBS prepayement
 
 
 
 
 

Senior Associate, Data Scientist

Capital One

Aug 2021 – Jan 2023 Mclean,VA

Capital One Financial Corporation is an American bank holding company specializing in credit cards, auto loans, banking, and savings accounts operated primarily in US.

Responsibilities:-

  • MBS prepayement
 
 
 
 
 

Data Scientist Intern

Ocrolus, Inc.

Jun 2021 – Aug 2021 New York City, New York

Ocrolus is a document automation platform that powers the digital lending ecosystem, automating credit decisions across fintech, mortgage, and banking.

Responsibilities:-

  • Worked with team to streamline data pipelines.
 
 
 
 
 

Data Scientist and Platform Development Intern

Lantern Pharma, Inc.

Jan 2021 – Apr 2021 Dallas,TX

Lantern Pharma is a uniquely positioned life science company pushing innovations in pharmaceutical biotechnology and drug-companion diagnostic co-development through Artificial Intelligence, Machine Learning, and Genomics.

Responsibilities:-

  • Worked with team to build component of Data Pipeline in cloud for Analysis of Cancer Drug exploration platform.
 
 
 
 
 

Data Scientist Intern

Capital One

Jun 2020 – Aug 2020 McLean,VA

Capital One Financial Corporation is an American bank holding company specializing in credit cards, auto loans, banking, and savings accounts operated primarily in US.

Responsibilities:-

  • Implemented parallel processing for tuning hyperparameters of model for prepayment rate prediction of Mortgage Based Securities(MBS) and reduced experimentation cycle by 30 mins via workflow optimization.

  • Built and analyzed automated 70 page performance report (PDP/ICE, and S-Curves) for comparing models and finalizing Hyper-parameters.

 
 
 
 
 

Machine Learning Specialist | Student Assistant

VHSS Lab at Center for Modeling and Simulation, UTD

Sep 2019 – May 2020 Richardson,TX

VHSS Lab, UTD is a cross-disciplinary team including full-time and student computer and behavioral scientists, human computer interaction experts, animators and modelers under Dr. Marjorie Zielke.

Responsibilities:-

  • Researched and trained transformer-based model for virtual patient interacting with medical students using Pytorch.
 
 
 
 
 

Artificial Intelligence Engineer

Huddl Enterprise Communication Pvt. Ltd

Apr 2018 – Jun 2019 Hyderabad,India

Huddl is the collaboration hub that brings the right people, information, and tools together to get work done.

Responsibilities:-

  • Involved in building components of products utilzing deep learning and natural language processing.
 
 
 
 
 

Associate Data Science Engineer

CoArtha Technosolution Pvt. Ltd.

Sep 2017 – Apr 2018 Hyderabad,India

CoArtha TechnoSolutions is an IT company and has a product called maprecruit.com which provides a parallel and intelligent recruiting platform.

Responsibilities:-

  • Engaged in developing feature of product for parsing job description and candidate filteration.
 
 
 
 
 

Associate Software Engineer

CoArtha Technosolution Pvt. Ltd.

Sep 2016 – Aug 2017 Hyderabad,India

CoArtha TechnoSolutions is an IT company and has a product called maprecruit.com which provides a parallel and intelligent recruiting platform.

Responsibilities:- - Part of Data Science team responsible for building ml capabilities such as matching algorithm for product.

 
 
 
 
 

Founder

DigiFledged

Jul 2015 – Aug 2016 Bhilai,India

Motivated by my previous training experience at DSIM, We started a small venture, Digifledged. Here, We started a small venture to help businesses to start their digital journey. We mainly served services such as web development, digital marketing, Analytics using machine learning, and excel automation/reporting projects services with macros.

Responsibilities:-

  • Built connections with clients and discuss requirements of projects regarding.
  • Developed & design website with excellent delivery on time with client satisfaction.
  • Developed & successfully managed a blog differencetalk.com and homeitarian.com with 130k+ pageviews between April 2016 to March 2019.
  • Delivered multiple projects and earned revenue of more than 200K INR within 1 year and became profitable.
  • Fulfilled more than 17+ freelancing projects on Excel dashboard and automation.
  • Level 1 Seller on Fiverr with more than 10, 5-star ratings.

Client Projects:- Projectcuriosity.in | Ophthalmicmart.com(old) | Aditispat.com | Ezoneservices.in

 
 
 
 
 

Junior Manager

JSW Steel Ltd.

Feb 2014 – Apr 2015 Karnataka,India

JSW Group is a $12 billion company and leading Indian multinational steel making company. It is one of the fastest growing companies in India with a global footprint in over 140 countries

Responsibilities:-

Managed a sub section of manufacturing area and provided analysis to upper management.

 
 
 
 
 

Trainee

Delhi School of Internet Marketing

Aug 2013 – Jan 2014 Delhi,India
Delhi School of Internet Marketing provides full-fledged live classroom training for Students industry executives and entrepreneurs. I attended the hands-on training is provided in batches in a structured where trainees have access to practical application opportunities to practice and apply learned skills.

Skills

ML algorithms

Linear/Logistic Regression, SVM, Clustering, Tree Algorithms, Naive Bayes, Ensemble Methods

Libraries

Numpy, Pandas, Matplotlib, Plotly, NLTK, Gensim, Sklearn, Spacy, Scipy, Tensorflow, PyTorch

DL algorithms

Feed Forward NN, RNN, CNN, Attention Mechanism

SQL/NOSQL

MySQL, MongoDB, ElasticSearch, DynamoDB, Neo4j

Tools and Technologies

Linux, Django, Rest-API, Docker, Kubernetes, ML-flow, AWS, Terraform, ECR, Hadoop, Hive, Spark

Langauage & Frameworks

Python, C, Java, Scala, Matlab, Django, Flask, Node.js

Projects

*
All Deep Learning NLP Machine Learning Web Development

Auto correct spelling and Auto complete using N-gram

Auto Correct is generally used to provide suggestion on mobile while typing and Auto Correct Spelling Preprocess data and compute word probabilities from corpus Generate words 1 and 2 edit distance away and filter based on vocabulary Suggest word with highest probabilities Auto complete is utilized to complete search query or while writing email.
Auto correct spelling and Auto complete using N-gram

Credit Card Transaction Fraud

Credit Card Transaction Fraud is a very common problem for banks and they use machine learning models to tackle this problem. Imputed missing data,created custom features,normalize and encode features Performed exploratory data analysis of features Handled imbalanced data using SMOTE and custom loss function.
Credit Card Transaction Fraud

Document search using approximate k-nearest neighbor

Preprocessed/encoded documents into vector for entire corpus Implemented local sensitive hashing(LSH) for multiple universe (different set of random planes) Developed document search using approximate k-nearest neighbor and LSH
Document search using approximate k-nearest neighbor

End-to-End Natural Language Generation for Conversational Agents

Dataset : E2E NLG Challenge Each instance consist of a dialogue act-based meaning representation (MR) and up to 5 references in natural language MR: name[The Eagle], eatType[coffee shop], food[French], priceRange[moderate], customerRating[3/5], area[riverside], kidsFriendly[yes], near[Burger King] NL:
End-to-End Natural Language Generation for Conversational Agents

Image Detection using CIFAR-10

Image detection is common task in computer vision.It is used in as a subtask for many important industrial application such as autonomous vehicle. Augmented image data using transformation technique(random crop,vertical flip) Implemented and trained RESNET family of architecture for image detection with 86% accuracy
Image Detection using CIFAR-10

Named Entity Recognition

Named Entity Recognition is a method to locate and identify important concepts within documents. I trained vanilla and GRU variant of recurrent neural network to identify person,location and organization from sentences.
Named Entity Recognition

Parts-of-Speech Tagging (POS) using Hidden Markov Model (HMM )and (MEMM)

Parts of speech tagging the process of assigning a part-of-speech tag (Noun, Verb, Adjective…) to each word in an input text. I have trained two different model HMM which is generative model and MEMM which is discriminative.
Parts-of-Speech Tagging (POS) using Hidden Markov Model (HMM )and (MEMM)

Question Answering using SQUAD 1.0

Language Understanding Capabilities of computers is crucial to many modern applications such as voice assistants,autocomplete suggestions for emails and answering questions by users in search engines.Out of them Reading Comprehension is an important and challenging problem.
Question Answering using SQUAD 1.0

Spark Streaming with Twitter and Kafka for sentiments of tweets on COVID-19 and India

Due to the second wave of covid-19, There is an unprecedented surge of covid-19 in india and we wanted to analyze the sentiment of public on the basis of tweeter. We have used stanford core NLP to analyze the tweets.
Spark Streaming with Twitter and Kafka for sentiments of tweets on COVID-19 and India

Twitter Sentiment Analysis with Deep Neural Network

Preprocess/encode tweets and create batch data generator for test/eval set Prepare word embedding for words for feeding into neural network Train and evaluate feed forward neural network using TRAX
Twitter Sentiment Analysis with Deep Neural Network

Web Based Contact Management

Simple web application to add,edit and delete contacts using model view controller using Django. Features Contact Display Window Contact Search Adding New Contacts Modifying / Deleting Existing Contacts
Web Based Contact Management

Word Embedding based on Continuous bag of words model

Implement cbow model to get word embedding with 2 different architecture First approach Bengio et al. neural language model Second approach Efficient Estimation of Word Representations in Vector Space Perform PCA on word vectors and visualize relation between few words
Word Embedding based on Continuous bag of words model

Accomplish­ments

Natural Language Processing with Probabilistic Models

Coursera Nov 2020
See certificate

Natural Language Processing with Classification and Vector Spaces

Coursera Sep 2020
See certificate

Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization

Coursera Sep 2020
See certificate

Architecting in AWS by The University of Texas at Dallas

The University of Texas at Dallas Jun 2020
See certificate

Neural Networks and Deep Learning

Coursera May 2020
See certificate

SQL

Linkedin Mar 2020
See certificate

Django

Linkedin Mar 2020
See certificate

Machine Learning

Coursera Feb 2020
See certificate

Pytorch

Linkedin Jan 2020
See certificate

Linked Data Engineering

Hasso Plattner Institute Dec 2016
See certificate

Contact