Ayush Agarwal Email & Phone Number

Seattle, Wa, Us

San Francisco, California, Us

1. Tech Lead [Next Best Action]- Spearheaded the development of an AI orchestration pipeline to help marketers identify their next best opportunity through a series of recommendations. Utilized a cutting-edge LLM API to dynamically generate message content variations along with determining a suitable retargeting audience for follow-up engagement.2. Mentored a data science intern as we created a model pipeline to serve rank predictions on subject lines to determine the best one for a campaign according to the send audience.3. Engineered scalable Spark transformations to clean and aggregate raw ad-event data, driving efficient loading into a high-performance time-windowed offline feature store with billions of weekly data points.4. Validated customer constructed predictive goals structurally, numerically and with respect to the history and drift of the customer data. This helped identify valid goals before they were trained with a XGBoost binary classification model to predict possible conversion numbers on them in the near future, saving hundreds of hours of GPU compute on goals that would have failed during training.5. Packaged all python utility code for various services in the data science repository into a lean and easily deployable wheel which has >95% test coverage. 6. Participated in a design sprint to progress towards a data product manager role in the team by identifying innovations that can help drive the Goals product forward to make it more actionable and explainable for our customers by surfacing audience and messaging insights collected on the conversion outcome of the goals.7. Winner of the “Best Reliability” Award at the Iterable 2021 Hackweek for our work to build a regular pipeline to garner and present metrics on Email Deliverability for different channels and send domains at Iterable. Also won the Iterable Red-Diff Challenge for making our code more sleek by deleting unneeded and non-production code in Iterable data science.

San Francisco, California, Us

1. Led the delivery of the patented brand affinity pipeline which assigns labels to all customers' users defining their relationship with brands. Helps over 3K brands to auto-segment billions of users into groups for specific campaign targeting.2. Contributed in 100x scale-up and optimization of data quality, model training and KPI pipelines, ensuring timely delivery of Send time Optimization and Brand Affinity in GA.3. Handled transactional and custom event data ingestion and built a model serving framework for the Product Recommendations pipeline.4. Improved data quality by managing a tiered delta lake controlling flow and hygiene of millions of daily data records for accurate feature generation. 5. Closely involved in codebase adaptability and regular documentation in accordance with advancements introduced in the tech stack.

Baltimore, Md, Us

1. I am working on my Master's Thesis at CIRL, JHU under Dr Gregory Hager. 2. My current work is on assembling, crowdsourced annotation and analysis of surgical video and kinematic data collected using the da Vinci surgical robot. 3. I will be looking to tackle the problem of Surgical Skill and Precision Evaluation by breaking it down into two subproblems - (a) the release of a new surgical training dataset called MISTIC-SL and (b) a comprehensive evaluation of some baseline and novel methods on the above dataset.

San Francisco, California, Us

1. Built Knowledge Graphs to extract Product Recommendations equipping clients with intelligent cross-channel marketing. Graph made with User-Product edges from product views and shopping funnel events.2. Constructed a robust pipeline using pySpark and Graphframes to handle a billion custom events and draw customer insights from user-product and product-product event interactions.

Baltimore, Md, Us

1. Analyzing radiological imaging using Deep learning algorithms (SSAE, CNN, DBNs) for detection of disease. Implementing and designing an interface for streaming algorithms using deep learning methods. 2. Designing a Generative Adversarial Network pipeline to support medical image synthesis using a limited available dataset. Conducted a comprehensive research and review of adversarial networks in the medical imaging domain.

Baltimore, Md, Us

Course Assistant for EN 601.415 Databases at JHU

Armonk, New York, Ny, Us

1. Project involved building a network to perform accurate hierarchical object detection in large images using attention focus mechanism in combination with deep neural networks.2. Used MATLAB for quadtree breakdown of image and Keras to calculate the latent space and the reconstruction of image by estimation of the joint probability. 3. Achieved results better than state of the art methods like RCNN, F-RCNN, YOLO.

New Delhi, In

UNDERGRADUATE THESISBrain MRI SegmentationADVISORS : DR. CHETAN ARORA (IIIT-DELHI) AND DR. ANUBHA GUPTA (IIIT - DELHI) • Detecting anomalies like schizophrenia, bipolar disorders, etc in the brain anatomy by doing semantic segmentation.• Segmentation performed by fully connected convolutional neural networks and other deepnets.• Concept devised to use a fan-in-fan-out network structure having skip connection for coarse and fine information sharing.• Decision to use the 3D structure of the MRI as a collection of slices: maintained a gradient between consecutive slices using optical flows to exploit the temporal structure.• Trained and Evaluated the model using various publicly available anatomy datasets to build a generalized system.

New Delhi, In

Prepared and organized practical labs for first year undergrad students. Also held tutorials and office hours for revision.

Master'S Degree, Computer Science

Bachelor Of Technology (Btech), Computer Science