Krishna Prasad

Experience

• 

  Societe Generale Global Solution Centre

  Aug 2018 - Sept 2022

  - Aided in budgeting through the development of statistical models consistent with US FED and ECB regulatory requirements to forecast Net Banking Income (NBI) across 15 Business Units (BU)- Experimented with time series (ARIMAX, Exponential Smoothening, VARMAX), cointegration(Engle-Granger, ARDL, ECM) and regression models (Cubic Splines, GAM). Included stress testingchecks to monitor model behavior during adverse scenarios- Ownership for statistical upgrade of NBI forecasting models belonging to Russian region (15 models) which involved understanding business activity, data collection, exploratory data analysis, feature selection, model development, forecasting results, presentation to clients and documentation- Developed, automated, and deployed a framework for periodic monitoring of model health to assess drastic shift in data distribution and accuracy to assist continuous monitoring of 103 models in production across all Business Units Show less - Aided in automating information extraction from Invoice documents and KYC forms using Optical Character Recognition. Used an ensemble of Tesseract V3 and LSTM-CNN model trained on IAM and RIMES dataset to handle digital text and handwritten words respectively- Responsible for building a pipeline to enhance the quality of documents for accurate extraction. Experimented with techniques such as Cycle GAN to denoise scanned documents using an unpaired internal dataset and SRGAN to enhance the clarity of low-resolution scanned documents- Assisted in document layout analysis, through semantic segmentation of the image using a pretrained DeepLab V3 model on PRIMA dataset. Developed a POC using Siamese networks for signature verification in forms- Developed REST APIs to serve the above features, integrated with User Interface, packaged using docker and deployed in OpenShift environment Show less

  • Data Scientist

    Apr 2020 - Sept 2022

  • Associate data scientist

    Aug 2018 - Apr 2020
• 

  Highavenue Creative Labs Pvt Ltd

  Aug 2020 - Jan 2021

  Machine Learning Engineer

  - Aided in the development of a mobile-based social media platform capable of generating lifelike 3D animation videos from photos uploaded by the user- Responsible for the development of a deep learning model to convert 2D images into 3D models. This included devising the problem, research on existing methods, model building andintegration with Unity3D for automated development of models- Initial approach was to perform semantic segmentation of human body parts, using a model based on cross-domain complementary learning followed by pose estimation and proportion calculation using rule-based methods. - Developed REST APIs to serve the model, packaged using docker, and deployed the container to Google Cloud Platform Show less

• 

  Fidelity Investments

  Sept 2022 - now

  Data Scientist