Parth Vadera Email & Phone Number

Sunnyvale, Ca, Us

Los Angeles, California, Us

Designed and implemented risk strategy architecture (block, trust, media network, and suspicious user mining) along with feature store and data pipelines from initial concept through to production, streamlined development processes, and led the recruitment and mentoring of new team members, significantly boosting team cohesion and knowledge exchange.Developed and deployed a real-time deep learning neural network fraud detection model at transaction level for financial payments. Managed the production rollout, including feature engineering and consistency checks, setting various alarms, and achieving significant cost savings and preventing $12 million in fraudulent transactions during the initial month of rigorous A/B testing. This in-house deep learning neural network model replaced third party solution Cybersource (by Visa) for blocking fraudulent transactions in real timeDeveloped a deep neural network model to generate user embeddings for segmentation, enabling tailored selectors and controls at the cluster level with customized strategies for each cluster.Working on next-generation recommendation engine using state-of-the-art transformer-based models. Employed techniques such as multi-modal learning, attention mechanisms, and contextual embeddings to enhance user personalization.

San Francisco, California, Us

Tech lead for developing global (America, Asia, Europe) forecasting capability at product level leveraged by planners and utilized as an input for other projects (Replenishment, Product Assortments, Pricing). Improved MAPE by 26 points across the board compared to existing rule based approach• Developed propensity models for Loyalty using wide and deep neural network model to address data sparsity (embeddings) and improved redemption rate across the board by 12 points YoY. Technologies Used: Tensorflow 2, Python• Refactored email targeting approach to account for the recent apple privacy updates which reduced the impact of this update on revenue/user.• Built a traffic forecasting engine to predict foot traffic at brick and mortar stores post reopening after covid shutdown for labor allocation and financial budgeting. Technologies Used: Python, XGBoost, Random Forest, DeepAR• Developed fit finder recommendation engine for similar products using embeddings from convolutional neural networks (Resnet50, Inception). Leveraged these embeddings to address cold start problem for forecasting sales of new products. Technologies used: Keras, Tensorflow• Built a what-if simulator tool to analyze the effects of changes in promotion price to overall sales for Levi’s brick and mortar business using probabilistic forecasting approach. Technologies Used: Python, DeepAR• Lead the pricing test initiative for Levi’s stores which includes store selection for experimental/control group, reporting, hindsight analysis, relationship with key stakeholders and working with cross functional teams.

Boston, Massachusetts, Us

Implemented attribution model to measure the impact of TV Ads on web traffic using synthetic control method, BSTS, and Long-Short Term Memory (LSTM). Technologies Used: R, Python, KerasDeveloped automation tool for exploratory data analysis of 3rd party data and successfully tested on Rentrak and comScore. Technologies Used: PySpark, AWS EC2

Boston, Ma, Us

Collaborated with Medical records data engineers from U.S. Department of Veterans Affairs to build generative classification models (bayesian belief network) in order to predict and understand relationship between the risk factors and mental health in US Veterans. Technologies Used: R (bnlearn, randomForest, kernlab, ggplot2) & SQLLed a team of engineers for an NSF funded project to predict the referral network in a health care system using Bayesian approach of combining evidence and network simulation. Technologies Used: Python (scikit, networkX)

Ahmedabad, Gujarat, In

Developed an optimized targeted marketing model for an e-commerce website selling traditional apparel using customer segmentation, predictive modeling and A/B testing. Technologies Used: R (caret, OAuth2.0) & Google Analytics.Developed a forecasting model to predict the number of daily visitors on e-commerce websites using Holt-Winters smoothing.Assisted with data extraction and feature engineering for building a recommendation engine using association rule mining for e-commerce websites.

Master Of Science, Operations Research

B.Tech, Information And Communication Technology

Ssc And Hsc, Life.