Education
-
- Present
University at Buffalo
MS in Data Science
● GPA: 3.8/4.0 ● Coursework: Statistical Data Mining, Data Model & Query Language, Probability Theory, Predictive Analytics, Machine Learning, Reinforcement Learning. ● AWS Certified Cloud Practitioner
-
- 2017
Indian Institute of Technology (IIT)
Bachelor of Technology
Achieved 99.76 Percentile in JEE (All India exam with 1.4 million candidates) for admission into IIT.
Experience
-
2022 - Present
The Perfect Child LLC,
Data Analyst Intern
● Built an interactive maps application to match clients and Therapists, decreasing time and complexity of matching.
● Automated repetitive data collection from a dynamic website with Selenium, saving 30 minutes daily.
● Achieved synchronization while maintaining data segregation by combining individual google sheets files with Appscript to synchronize changes made in both the combined and individual sheets.
● Replacing a buggy software for logging therapy data, manually calculating billing hours for therapists by building a streamlined webapp hosted on AWS saving $5k/month for a small business, increasing efficiency, and gaining complete access to data. -
2019 - 2021
Merilytics
Associate Data Scientist
● Engineered an Automated Valuation Model (AVM) using TensorFlow-based custom nearest-neighbor architecture to identify comparable properties by enriching core dataset from multiple sources, and lead a team of 3
o Automated property valuation resulting in ~$25 million annual savings by decreasing ~2700 man hours weekly.
o Built the SQL ETL pipeline to extract features for real estate properties from ~100 GB of OSM US maps data.
o Collaborated with all stakeholders on biweekly calls to deploy the AVM on AWS SageMaker.
o Reported iterative improvements, and insights for adding variables on client facing weekly calls.
o Established Quality control checkpoints in the data pipeline prevent data inconsistencies and other issues.
● Developed a Demand forecast model with 1800 SKUs for a major European EV supplier.
o Leveraged Time Series forecasting packages such as fb-prophet, neural prophet, & seq2seq neural nets.
o Reduced time taken for forecasting from 1 week to 4 hours with an end to end pipeline.
● Optimized scheduling with a heuristic-based driver scheduling algorithm to incorporate business constraints.
o Produced a simulation of bills movement across 34 terminals for a long-haul trucking client.
o Removed human input for driver scheduling, optimizing time taken from multiple days to 15 minutes.
● Built Sales Forecast Model using Keras for creating business strategy for an American online clothing chain.
● Developed an end-to-end data pipeline on Azure utilizing Azure functions for daily scheduled ETL.