Sana Basharat

-Implemented ensemble learning models to predict long performance test outcomes, cutting testing times by 30%, automating operations and reducing factory energy consumption leading to a World Economic Forum award for the factory -Built time-series models for sensor data belonging to more than 275K products to predict failures -Designed web scrapers using selenium and BeautifulSoup to collect customer reviews and present a thorough brand analysis of Whirlpool Europe leading to a successful €38M acquisition

- Built ETL pipelines using Python, DBT, Airflow, and Fivetran, leveraging AWS and GCP for multiple retail and health-tech clients. - Designed an enterprise-level data ingestion and transformation architecture for our in-house data-driven marketing analytics platform

- Wrote complex SQL KPIs to track Electronic Patient Reported Outcomes for more than 12 clinics associated with a digital health startup. - Implemented ETL pipelines to pull content performance, and audience behavior data from Facebook, Shopify, Magento, WooCommerce, Google Ads and Google Analytics for multiple retail brands. - Made predictive time-series models for revenue and cost prediction with an error rate of less than 10%.

Worked with a US-based client on a property management system (PMS), using NodeJS/Express, VueJS, MySQL, Docker, K8s and AWS.

An LA-based AI-generated synthetic media company. Primary roles: Writing technical content and blog posts Managing social media accounts Documentation

Tasked with modeling various business problems with AI-based solutions, we were given a lot of flexibility over the domain we wanted to work in, and how we wanted to tackle the problem. For the following projects, I used Python and its libraries like OpenCV, Scikit Learn, LIME, SHAP, Keras, etc. Projects: Breast Cancer Prediction using ML, explained using LIME+SHAP Stock Price Prediction using LSTMs, explained using LIME Melanoma Classification using CNNs Anomaly Detection for Insurance Fraud

- Developed a tool to classify non-coding mutations as cancer drivers or non-drivers - Designed a feature selection and feature engineering pipeline for 70 annotated variant features and implemented XGBoost models for driver detection

GPA: 3.30/4.00 Final project: Video-based Automated Cricket Commentary Generation