Dec 2024 – Nov 2025 · 11 mos · Kolkata, West Bengal, India · Hybrid

• Managing Projects with Generative AI use cases.

Apr 2023 – Dec 2024 · 1 yr 8 mos · Kolkata · Hybrid

• Model Validation and Risk Assessment for Retail Lending Products
• Validated ML Models and their Use-case with Statistical Techniques – Kolmogorov Smirnoff (KS), Area Under the Curve (AUC) and Gini Coefficients for Out of Sample (OOS) data and Rank Order Flips, Population Stability Index (PSI) and Characteristics Stability index (CSI) for out of time (OOT) data.
• Carried out ML Model Monitoring and ML Model Retraining on a periodic basis
• Framed AI-ML Validation Policies for the Bank on Gen AI related Use Cases.
• Used Terraform to set up AI-ML resource for Team Members in Microsoft Azure
• Worked on Identification of Risks in Collection and Income Estimation Models, Auto Loan Default Model and Bureau Default Models
• Created and tested Challenger Models in Azure by creating Data Assets, running Command Jobs for Model Training, Performing Hyperparameter Tuning and registering the Model with the best metrics score derived from MLFlow and used CI/CD for Training and Deployment through Azure Devops.

Dec 2021 – Apr 2023 · 1 yr 4 mos · Bengaluru, Karnataka, India

• Worked on Predictive Modelling on Logs and Health of Load Balancers. Used AWS, Deep Learning and Statistical Techniques for Inferential Analysis.

Apr 2021 – Dec 2021 · 8 mos · Bengaluru, Karnataka, India

• Conducted Model Analysis for Multi Label Sentence Classification. Used GCP AI Notebook and Cloud Storage for Data Fetching. Prepared Plots for Clients' understanding on the present performance of models.

Oct 2020 – Apr 2021 · 6 mos

• Implementing Customer Cancellation Prediction of their product post purchase.
• Working in Google Cloud Platform for Model Creation and Production.
• Used Weight of Evidence, Information Value approach for Feature Selection.
• Used a mix of Linear Sequential as well as bi-directional LSTM/GRU oriented approach for model training. Improved client’s hackathon’s accuracy from 76% to 92.5%.
• Used Models Predictive analytics to arrive at Descriptive Analysis over order life cycles.

Jan 2020 – Oct 2020 · 9 mos

• Worked on Speech to Text Assignment for a Healthcare Client with the use of Linux, Python and Deep Learning.
• Implemented Mozilla's Deep Speech custom Speech to Text Model for Call Audit purpose.
• Used ML algorithm - Naive Bayes Classifier to classify Call Purpose and Call Subcategory .
• Used Python and Spacy to create Custom Entity model to train and later infer Business Entities discussed in the call.
• Used Pyaudio Analysis and Fuzzywozzy to diarize Agents voice and later infer Sentiment of Call and derive the profile of agent through Text Blob.
• For Production level deployment Used Flask and Nginx through Docker-Compose to create REST API.
• Used Mongodb in Docker Image, to store Information of the scripts which was later passed through the ML algorithms and then passed to the client as part of POST requests.
• Used both Windows and Linux extensively in developing and later in deployment.

Jul 2019 – Jan 2020 · 6 mos

• A. Implemented Store Sales Forecasting for various Stores, Departments and Categories by taking into account all the relevant features like weather, holidays, lags, shift etc.
• B. Used Time Series Analysis using SARIMAX and Neural Networks - LSTM and Dense in Keras and Python for training and Building Model.
• C. Implemented XGBoost and LGBM as a technique for comparing the effectiveness of the model.

May 2019 – Jul 2019 · 2 mos

• Project: Amgen (Life Science Domain) || Role: Data Lake Analyst:
• A. Worked on Amazon Comprehend Medical with Pyspark in Data Bricks Environment for an NLP based assignment, by extracting meaningful information w.r.t. Patients and their relevant entities. Used Python Libraries Spacy, Nltk, Pandas along with SQL.
• B. Post the Creation of data, transferred them into Amazon Red Shift environment from where used Tableau for Data Visualization for count of each entities by preparing a Word Cloud. Used Parameters and Calculated Fields along with Filters and Dashboards.
• C. Understood scheduling task with Airflow’s Directed Acyclic Graph (DAG)

Jul 2016 – May 2019 · 2 yrs 10 mos

• o Elicited and Prioritized Product Backlog for various high-level functionalities.
• o Did Time Series Forecasting on Revenues for Zones and Sub divisions for Secure Linked Meters.
• o Did Multi Class Classification of Customer queries (NLP) related to Meter Burnt, Low Voltage, High Consumption, Voucher Problem through Random Forest Classifier and later through GRUs.
• o Performed Analysis via Web scrapping with Python Pandas for retrieving AT&C losses of various Utilities and comparing with WBSEDCL through visualization graphs Seaborn, Plotly and Matplotlib.
• o Prepared Functional Design Documents with SMEs by taking approval from Key Stakeholders.
• o Prepared Go Live Plan by entailing all the processes and functionalities.
• o Reviewed Test Cases to ensure Quality Criteria is met by Conducting various Iterations and UATs.
• o Analyzed Solutions Limitations Post Release through Exploratory Data Analysis
• o Documents Prepared/ Assisted include: Business Process Modelling, Requirements, Solution Requirements, Use Case, Iteration Planning, Design, Pre-Release User Manuals, BACCM, etc.

Aug 2011 – Jul 2014 · 2 yrs 11 mos · Hyderabad Area, India

• Project: Highmark (Health Care Domain) || Role : Data Warehouse Testing
• Preparation of Test Cases from Requirements for Claims Data of Customers. Creation of Test cases from the requirements and building the respective SQL queries for each test case and thereby Executing those Test Cases by running SQL Queries in Teradata and validating the mapped Data, Raising Defects for the incorrect mapping of Data.
• Project: Chrysler (Automobile Domain) || Role : Regression Testing
• Execution of Regression Test Cases in SAP Solution Manager. Lodging Defects for those Phase 2 Regression Testing and Closing down of Defects post provision of the correct Data.