Neha Joshi Email & Phone Number

United States

Framingham, Massachusetts, United States

- Built a Flask-based REST API to automate the retrieval of LSTM and BERT-based models for pricing. - Performed performance benchmarking of the Fast APIs using caching, and multithreading, optimizing I/O operations, boosting the performance by 99.95%.- Optimized price forecasting code on Databricks by leveraging Apache Spark, reducing data processing time by 90%

Boston, Massachusetts, United States

TA for undergrad course CS3000 : Algorithms and Data

Boston, Massachusetts, United States

TA for undergrad course CS3000 : Algorithms and Data under Prof. Elena Strange.Responsiblites included conducting office hours and helping students through their queries.Also conducted interview style questions' black boarding recitation sessions with the students.

Pune, Maharashtra, India

- Developed and maintained an ETL pipeline in Jenkins for 50+ microservices based on Apache Spark, Python, and Scala, handling and transforming 1 trillion+ data points.- Designed and built scalable REST APIs in Django and MongoDB to support price edits by the end user to update the predicted product prices to 3TB+ data.- Developed robust data migration pipelines via shell scripting to facilitate the seamless migration of reporting PostgreSQL database module data from Amazon Web Services to Azure Cloud, guaranteeing zero downtime and achieving a remarkable 90% reduction in data management expenses.

Pune, Maharashtra, India

- Constructed a cross-system automation framework written in Java, based upon Java Messaging Service, Java Spring, Junit 5, and Selenium which reduces the time to validate testing workflow manually by 80 %.- Reduced the REST API response time by 20% by caching repeated request calls and removing redundant calls to other microservices by using Redis as a cache.- Developed a Python-based analytical tool utilizing real-time streaming of user log data through Apache Kafka, facilitating in-depth analysis of internal platform usage and leading to the decommissioning of 4 on-premise servers, resulting in cost savings.

- Utilized input data to design a Decision Tree algorithm, achieving 98% accuracy in predicting Huddling actions with refined rule sets.- Achieved 93% accuracy by developing an Image Segmentation-based algorithm for human hair color prediction, with a customized training and testing pipeline in Google Colab.- Reduced 100% of manual data collection tasks by designing and building data ingestion pipelines.

Pune Area, India

Master'S Degree, Computer Science

Bachelor Of Technology - Btech, Electrical, Electronics And Communications Engineering