Rishav Das

Responsible for performing research in Readmission Risk modelling and guiding the Data Scientist based upon business needs. Responsible for pulling data from different API feeds ingested to redshift, preparing datasets, performing EDA, Feature engineering, and feature selection. Executed classical models and applied hyper-tuning techniques. Designed and developed Attention based LSTM model for learning sequence series. Achieved AUC of 0.86 and accuracy of 0.89.

Responsible for end to end machine learning project development and architecting the life cycle of the system using MLOps. Developed conversational AI platform for diversified language (Japanese/Chinese). Responsible for designing and developing multi-lingual BERT for Japanese/Chinese from English. Overall accuracy achieved was 92.84%. Responsible for embedding the deep learning model into docker for deployment with Salesforce CRM platform. Filed 15+ patents.

Responsible for performing research in disease progression using graph neural networks and developing causal inference model for determining the disease cause. Project involves 2.7 million patient records with 120+ features. Objective is to determine disease progression for Type2Diabetes and other chronic disease. Model is in research state, working on heterogeneous KG, bayseian network, counterfactual, and referring biostat Algos.

Responsible for developing language model using Attention based Bi-directional LSTM model for analyzing patterns from Cloud & Storage server logs. Developed Attn-Bi-LSTM to analyze logs without human intervention. Responsible for managing junior engineers on text pre-processing and data flow API integration. NLU model was productionized with accuracy of 92.67%. Worked on email classifier based upon project based NER.

Responsible for developing medical Imaging solutions to improve fracture Xray report scans. Used transformation & smoothing techniques and developed MASK RCNN for localizing the fracture and effected region. Model achieved an accuracy of 89.12%. Filed 20+ patents.

Developed RPA based intelligence for infrastructure bot using Optical Character recognition, Pattern detection, and ELK stacks for superfast log analysis. Filed 6 patents while working on infra automation research.