Kushal Dave Email & Phone Number

Sunnyvale, Ca, Us

Redmond, Washington, Us

Lead team of applied scientists by working on state-of art-techniques the ads search scenarioon technologies like building sparse neural network models for traditional information retrieval,NLU models based on BERT/DeBERTa to generate representation used for retrievals usingANNs, Natural Language Generation models distilled from LLMs to use it in retrieval scenarios. Leading several projects on search for Bing Ads like query representation, ad representationusing most recent models developed by my team in Natural Language Understanding (NLU)and Natural Language Generation (NLG) models.Work on shipping small LMs distilled from LLMs like LLaMA, Phi2.5, GPT4 models onproduction online.

Redmond, Washington, Us

Working on Deep Learning, Information Retrieval, Extreme classification and Nearest neighborhood search techniques to deliver high impact NLP and ML solutions to the core problem of ads selection for Bing Queries.

Redmond, Washington, Us

Redmond, Washington, Us

Predominantly working on online semantic techniques to retrieve relevant ads for queries.

Redmond, Washington, Us

Hyderabad , Telangana, In

Hyderabad , Telangana, In

· Enterprise Search: Building an Enterprise Search system for CSIRO corpus. This includes role based search to restrict access to organization's data based on user privileges, Semantic Indexing of documents using domain specific ontology, search on named entities, navigation through related documents using `Automatic hyperlinking' between documents etc. · Address Validation Engine: This project involved building a system for validating user en- tered addresses on Amazon's shopping portal and suggesting corrections for mis-typed address using a predefined Address Database. The system uses HMM models to suggest correction for the mis-typed addresses and used Lucene for quick retrieval of suggested corrections.

Sunnyvale, Ca, Us

· Worked on social targeting project for advertisements. The idea is to target user's who can incite others to click on the Yahoo! ads. Yahoo! messenger graph was used for this purpose. The ads shown on Yahoo! properties are social, i.e they have the feature of sharing it on Facebook, Twitter, Buzz etc. · Worked on analyzing the influence on Flickr social network. The work involved analyzing the role of the user, user's social neighborhood and the particular information to be propagated in the network. We also built a regression model to predict the depth of propagations for a user and particular action (information). The work is under review at a conference. · Worked on few Ad-hoc projects. One project dealt with analyzing the click through rate (CTR) of friends in a social networks. It was found that friends tend to have similar CTR values. Other projects include working on various definitions of information propagations, such as `joining a group', `marking an image as favorite'. It was found that different actions have different propagation dynamics.

Pune, Maharashtra, In

1. Worked on designing Web portal for AccessData Inc. US. · Involved in developing AJAX scripts for the web pages. · Technology: VB.NET, AJAX, JavaScript 2. Simulation of networks for assessing the performance of Motorola's WMS tool. · Involved in developing simulated networks for the WMS tool and exploring the perfor- mance of WMS on various kind of networks. · Received client appreciations for automating the data collection process. · Technology: Java, Perl

Doctor Of Philosophy (Phd), Information Retrieval

Education record