Nitish Kumar Ranjan

Experience

• 

  Biofest, IIT Madras

  Feb 2014 - Apr 2014

  Web Development

   Worked as a coordinator in the frontend Development in a team of 5 members in Biofest Implemented frontend animation using webkit and animation in CSS3 and was tested on various web browser including Google Chrome, Mozilla Firefox and Edge Got appraisal from the Lecturer Dr Shannon Olsson and other professors for the website

• 

  National Service Scheme

  Aug 2014 - Aug 2015

  Project Representative

   Represented a project namely “Wikipedia translation in Hindi” striving to help students for deeper understanding of the knowledge of subjects in Hindi Led a team of 8 members to create a highly effective translation of Wikipedia articles

• 

  Shaastra, IIT Madras

  May 2015 - Nov 2015

  Web Development

   Worked in a team of 8 to fully develop the frontend part of the Shaastra website using AngularJs Responsible for the User-Authentication via login-registration part of the website

• 

  Bharti Script, Indian Institute of Technology, Madras

  Jan 2017 - Mar 2017

  Web Development

   Bharti Script is proposed by Prof. Srinivas Chakravarthy as a unifying script for all Indian languages Designed and Developed user- friendly website that reduced the loading time of the landing page by using AngularJs framework

• 

  Self

  Feb 2017 - Jun 2017

  Android Development

   Conceptualized, designed, developed, and deployed the Android application with a graphically interactive User Interface The app describes multiple tourist places with pictures along with a short description of 5 major districts in Bihar Designed the features of the interface along with application components and UI resources

• 

  GyanData Pvt Ltd

  May 2017 - Jul 2017

  Data Analyst

   Assessed the healthcare datasets of speciality clinics having 248.5k patients, using R language in Rstudio Applied machine learning algorithms such as linear regression, clustering and decision tree on the dataset for deeper insights and devised solutions of multiple what-if problems Optimized revenue and expenses of healthcare dataset by undergoing detailed study using graphs and figures

• 

  DeTect Technologies

  Jun 2017 - Jul 2017

  Software Development

   Worked with the Three.js library which uses WebGL to display animated 3D computer graphics in the web browser Developed a web-based software module as a part of the product Noctua that would be helpful in visualizing the gas leakage in industrial pipelines

• 

  Vivriti Capital

  Jul 2018 - Apr 2020

  Analyst - Data Scientist

  > Term Loan Pricing Model- Developed Term Loan Pricing Model algorithm to provide pricing recommendation for the investors.- Involved Data Cleaning, Data Imputation and feature engineering of the dataset.- Applied 3 Sampling methods (Normal, Stratified and Cross Validation) to sample data and implemented Random Forest and XGBoost model on the dataset.> Data Standardization using Excel VBA (Macros)- Used Excel VBA to standardized various datasets including Mutual Fund Datasets (Multiple Fund House), Portfolio Cuts, Funding Sheet (Borrower’s profile) and MCA Data (charges Data)- The standardization was done on either monthly or quarterly basis or whenever needed. Show less

• 

  Yubi

  Apr 2020 - Mar 2023

  > Transaction Scoring Model- Developed a model for scoring the transactions that took place on the platform (CredAvenue) on a monthly basis.- Model include transaction parameters like Credit Enhancement, Collection Efficiency, Current Rating, Rating change, etc. - Implemented regression and the weighted sum of the parameters on the dataset to achieve the desired output.> Client Recommendation- Made a recommendation algorithm that recommends to the investor basis their preference like Revenue, Rating, EBITDA, PAT, financial ratios, sector in which they would like to invest, company vintage and profitability, etc.- Used several data sources like client financial data, Charges Data, customer data pertaining to either redshift or mongodb apart from the investor preference data to generate the recommendation score. Show less

  • Senior Data Scientist

    Aug 2022 - Mar 2023

  • Associate - Data Scientist

    Apr 2020 - Aug 2022
• 

  Spocto - A Yubi Company

  Apr 2023 - now

  > Next Best Action Model (Recommendation Model)- Applied LSTM Model (along with Functional API) to predict the suitable medium, time, message content and language to connect to borrower so that they have higher propensity to pay.- The model uses features derived from current running campaign, last month campaign and CIP (Customer Input Parameter) data to generate a input sequence to the model.- Also, used RL (Reinforcement Learning) approach to get the most suitable journey to approach the borrower. > Loan Propensity Prediction for SPOCTO- Used AutoML (EvalML & H2O) on the pre-due credit card dataset provided by the SPOCTO team to predict the propensity for the borrowers to repay the loan.- Extensive EDA was done on the dataset to select and engineer the features used for the model.> Sentiment Analysis- Performed Sentiment Analysis on the given dataset into four segment namely positive, negative, neutral and mixed- Performed NLP based Tokenization, Lemmatization, vectorization and processed data for it to get processed by the ML model- Applied TF-IDF and modelled with XGBoost algorithm to determine sentiment polarity of the dataset. Show less

  • Lead Data Scientist

    Apr 2024 - now

  • Senior Data Scientist

    Apr 2023 - now