Pranav Bansal Email & Phone Number

Seattle, Wa, Us

~ Engaged in Explainable AI research on Vision Transformers with 12 attention layers, reducing model bias and adversarial prompting risks by 30% through targeted interventions.~ Applied ML techniques (LogitLens, AttentionLens, and TunedLens) to identify the top classes within intermediate layers and Generated heatmaps using Layer-wise Relevance Propagation (LRP), improving model interpretability by 25%~ Increased model transparency and reduced adversarial vulnerabilities by 35% through systematic bias detection, verified using tests like the SHAP score, adversarial robustness tests, and fairness metrics

Working as Teaching Assistant for History in Computation undergraduate course, which involve teaching discussions, evaluating exam papers and projects and conducting office hours.

~ Engineered ETL Data Pipelines using Google BigQuery, Teradata and Amazon S3, to generate data warehouse, later produced interactive dashboards for retail store inventory and sales reports, reducing manual effort by 4 hours. ~ Optimized already existing ETL jobs for Big Data Processing using Hadoop and Spark for scalable data pattern and execution time, resulting in 30% reduction in latency for orders report creation and loading time.~ Designed and deployed Computer Vision AI model for object and action tracking in real time, optimizing accuracy by 25% by incorporating features from IoT device inputs including weighing scale, infrared and pressure sensors.

India

~ Developed ML Models: SIR, ARIMA and LSTM based, to forecast COVID-19 Cases sourcing data from the India-level time series dataset, consisting of 60M records. ~ Conducted a comparative data analysis of the various ML models, determining that SIR (Susceptible-Infected-Removed) achieved the highest accuracy of 97.6%. ~ Executed frontend and backend for a website, consisting of COVID-19 cases forecast graph for 28 Indian states build graph-based data model and query pattern using GraphQL APIs and backend using NodeJS, while front end build with React.

Master Of Science - Ms, Computer Science, 3.7/4

Bachelor Of Technology - Btech, Computer Science, 9.4/10

Education record