Sachin Varriar

Simplified loan application for MSME loans. Deployed model as RESTful API using FastAPI. Designed Data warehouse for updating model using Bigquery and Airbyte. Scheduled daily user scoring using Apache Airflow. Designed a data pipeline using python and airflow. Segregated PII data and user financial data into separate environments. Built Business Rule Engine tool for loan underwriting.

Taught SY306C - Introductory Statistics with R and NLP with Python and SQL as part of the MBA program in Amrita Vishwa Vidhyapeetham.

Took the lead in setting up the Data team at Vauld in both strategy and implementation. Setup Data warehouse in Snowflake to work with a modern data stack, with data ops using github, dbt. Integrated front end, backend and external data using segment, stitch and snowflake. Defined data quality checks for data pipeline and implemented the checks using dbt cloud. Ran multiple product related analysis based on frontend and backend events data. Built dashboard for Marketing performance. Liaised with directors, product managers to develop feature specific data marts.

Led a team to model key-value pair extraction for variables of interest for knowledge work automation. Deployed a ML model to for classify different type of documents using a simple Naive bayes classifier. Implemented Flair Spanish NER as a replacement for Spacy to improve capturing of Company names, Board of directors, Attorneys, Notaries etc. PoC using a T5 translator followed by siamese networks on Spanish documents to understand semantic similarity for legal terminology.

Maintained and tuned propensity model to determine potential purchase from dormant customers of CAT construction machinery. Played a key role in maintaining the quality of the data pipeline. Created SAS scripts for data quality. Created a feedback loop between Data Engineers and Data Scientists to report data quality issues. Predicted return of dormant customers back to CAT business with 53% conversion rate using Propensity modeling. Built a data pipeline using bash scripts scheduled on Azure high performance computing.

KPI tracking Dashboard for Invoice of Indices to media and brokers developed for stock exchange in the middle east. Worked on a PoC for Ontario Ministry of Government & Consumer services, Dashboard to help identify Consumer issues, Sentiment analysis and fraudulent market on Twitter and Yelp data. Managed Data warehouse services for a major Piping solutions company.

Developed Random Tree based model to decide optimal means of shipping of repaired parts for a major Turbine manufacturer. Used previous shipment data and data on contracted carrier rates to build models to predict optimal shipping methods.