Created baseline XGBoost models for Payment Risk and Pre-payment Risk and refined both models with additional feature engineering, feature selection techniques like backward stepwise feature elimination, PDP plots, PSI thresholds, IV, VIF etc. Created baseline linear, lightGBM models as well to test performance against XGBoost.Developed an end-to-end framework to rapidly iterate and create these models for other clients. This reduced time taken to replicate similar models for other clients by 60%Created a framework to evaluate different steps of MLOps lifecycle - Development, Deployment, Scoring and Governance. Benchmarked performance metrics at each stage for all the Azure services used in the flow.Developed model governance framework with metrics, monitoring framework and stability thresholds by analyzing at model performance over several vintages/OOT datasets, stress testing inference etc.Streamlined and automated governance metrics generation and reporting processes through python scripts. Deployed scripts on event-driven Azure Functions to automate metric generation as per defined cadence.

Developed a behaviour score to monitor the performance of customers to be used in portfolio management. Created more than 3000 features on data sourced from multiple channels(internal, Cibil, Equifax), and finally re- duced it to 45 varibales using bivariate anaylsis, information value and recursive feature elimination.Built a logistic regression model to get propensity to default for each customer, rank ordering customer segments into deciles. Model was tuned monthly with internal data and quarterly with bureau data.Lead a team of 2 Data Engineers to create a loan management system as data had to migrated to an analytical db from it’s current location. This facilitated faster tracking and reporting of key business metrics, it also automated a lot of tasks that required substantial man hours every month.Set up dashboards for daily TTD population monitoring, daily and monthly collections performance which helped streamline many business activities. These dashboards were created for the founders and heads of departments.Assisted in portfolio management activities, performed in-depth analysis and made business recommendations on adhoc basis.

Developed a Data Pipeline using SQL Data Extraction to filter out vast amounts of data and extract relevant features.Carried out feature extraction and transformation on a news dataset using CountVectorizer and tfidfTransformer.Built Naive Bayes and Support Vector Classifiers and optimized their performance using Stemming and Lemmatization. Carried out hyperparameter tuning for the classifiers using GridSearch achieving 91% accuracy.Leveraged Java implementation of Stanford's Named Entity Recognizer for POS Tagging, Word Tokenizing and carried out Entity-Relationship extraction on the same news article dataset.Analyzed current drug market of 5 major European countries for an over the counter(OTC) drug.Researched the market to find out key stakeholders in patient journey, from awareness of symptoms to fulfillment.Designed exhaustive surveys for patients, doctors and pharmacists to find out statistically significant factors affecting all the stages of patient's journey, thus drug's sales.Applied investigative skills and analytic methods to assess business impacts and make recommendations over the drug's marketing policies, required sales support.Worked on multiple chat-bots, automating the customer support and employee assistance apps for the client.Developed bots on Microsoft Azure, IBM Watson using their NLP Engines along with inter-communicating Node servers, Javascript and MongoDB. Deployed bots on-premise and on Azure cloud along with their dependencies.Developed Node scripts to pre-process huge amounts of data, filter out relevant text and load it onto MongoDB.

Worked in a team that was responsible for analysis and reporting regfarding commisioning of a Sulphur Recovery Block at IOCL Haldia, to be operable by 2018.Made reports on Claus, LTGT and Amine Regeneration processes to be used in the new block, reporting the feasibility and efficiency of the processes.

Webczar is a group of Designers and Developers, that provided services such as website design to development, mobile application development and graphic designing. Worked as a Web Developer and was able to complete many projects successfully with the team. During my stay there I was responsible for coordinating and leading the front-end developers and developing the back-end requirements of the projects.