Feb 2020 – Dec 2021 · 1 yr 10 mos · San Francisco Bay Area

• Building the AI powered, voice enabled digital assistant for doctors. Development of Suki speech service (S3) -- core AI + ASR engine powering Suki,
• Responsible for all things related to Machine learning at Suki
• Key highlights include:
• Spoken commands to intents (99.5% accuracy)-Highly scalable system with minimal inference latency <10 microseconds.
• Enhanced functionality via addition of commands and features

Aug 2018 – Feb 2020 · 1 yr 6 mos · San Francisco Bay Area

• Led the NLP/NLU efforts for Content and Commerce AI. Responsible for creating and integrating Content Intelligence solutions as a part of the Adobe Experience Cloud B2B solution suite. Conceptualized and developed NLP solutions for Content Management, Campaign execution, Analytics and Commerce product lines.
• Developed document classification pipeline allowing document tagging along various attributes (95+% accuracy) for content management. Capable of handling 2M+ document titles per day.
• Created automated clustering pipelines for identifying - primary issues from text fields in feedback forms, ticket issues from tech support conversations. Capable of handling 20K+ forms/tickets per day. Outlier detection using clustering results.
• Designed and developed a chat temperature monitoring system for assisting customer service representatives and identifying escalations. Capable of handling ~100K chat conversations.
• One of the six projects selected for Adobe’s popular “Summit Sneaks”

Mar 2017 – Aug 2018 · 1 yr 5 mos · Palo Alto, California

• Led the development of ML based fraud detection and NLP enabled solutions for user generated textual content (reviews, ratings, questions, answers) on Flipkart, .
• Developed the user review moderation pipeline with shallow network for classifying reviews (99% accuracy), question moderation pipeline (96.5% accuracy) and answer moderation pipeline (94% accuracy) with ensemble of shallow networks. Currently deployed and capable of handling 10M + reviews, 1.5M + questions per day.
• Created entity extraction pipeline for extracting specific entities from texts using sequence to sequence network architectures using bidirectional LSTM (98% accuracy). Currently deployed and capable of handling 2M+ requests per day.
• Designed and developed question deduplication pipeline using LSTM network with concatenation. Currently being evaluated for deployment (81% accuracy).
• Implemented shallow network model for quick sentiment detection pipeline with 99% accuracy on polarity of the sentiment.
• Created fast auto titling solution utilizing POS tagging and pattern matching.
• Winner of Flipkart Business Excellence Award

Apr 2015 – Mar 2017 · 1 yr 11 mos · San Francisco Bay Area

• Understand requirements, research and develop state of the art machine learning solutions for internal and external customers. Actively involved in developing all parts of data pipeline including ETL, feature engineering, modeling, data visualization, and product-ionizing.
• Designed and developed a context aware model (with 70% recall rate) using multi-armed bandit algorithm to predict the apps used in the following hour on a user’s mobile phone based on the current time, location and historical usage. The choice of algorithm allowed the model to automatically learn changes in behavioral patterns.
• Conducted research on policy gradient methods, developed and published a K-fold method for baseline estimation in policy gradient algorithms.
• Developed classification models using deep learning to classify real estate listing images based on certain characteristics (e.g. presence of granite counter tops, fireplace etc.), with 90+% accuracy.
• Leveraged mobile usage data to understand the key drivers of battery drain in Samsung Galaxy S6. Developed several metrics to provide actionable insights by classifying users and characterizing charging/discharging patterns based on battery consumption. Developed a regularized linear regression model to quantify the battery consumption for various apps, activities and behaviors.
• Other projects include: using deep learning for classifying distracted drivers (part of Kaggle competition), clustering of users based on app usage, and several descriptive analytics projects for understanding the behavior of mobile phone users.

Aug 2011 – Apr 2015 · 3 yrs 8 mos · Washington D.C. Metro Area

• Research Program Development, Multi-physics Modeling and Experimentation.
• Developed the high fidelity computational model of the human head, executed tasks include calibration of material properties via optimization techniques, validation of model, quantitative characterization of folds in the brain, statistical analysis of the effect of folds, and development of post processing strategies for analysis and visualization.
• Created novel corrosion pit growth simulations by incorporating the effect of locally varying crystallographic properties during the evolution of growth front.
• Awarded Alan Berman Research Publication Award

Aug 2006 – Aug 2011 · 5 yrs

• Doctoral research: Theoretical and experimental analysis of the effects of crystallographic anisotropy during micro machining.
• • Developed crystal plasticity based theoretical models for predicting the variation in cutting force with crystallographic orientation.
• • Designed and constructed two micro machining set-ups for performing micro-scale experiments to validate the theoretical models.
• • Formulated experimental techniques for machine and tool characterization, and sub surface deformation measurement.
• Applied research: Practical applications of micro machining.
• • Designed a multi-stage approach using mechanical micro milling and soft lithography to fabricate micro-fluidic channels with 3-D circular cross-sectional geometries.
• • Designed and developed a split die using precision engineering and micro machining techniques, for performing Equal Channel Angular Extrusion experiments.

Jun 2005 – Jul 2006 · 1 yr 1 mo

• Process modeling
• Developed computational tools aiding in quality control of display films.
• Performed finite element analysis, including nonlinear analysis for large deformation
• of polymers, thermo-electrical and thermo-mechanical simulations.