Generative AI capabilities for Shorts creation

Veo, Imagen, Gemini models

Video generation post-training

Google Discover: Content recommendation, personalized image recommendation, content understanding.

Content understanding, ranking, recommendation

Google Search

Built a multi-modal NN video page classifier.

• Replicated state-of-the-art techniques for the Semantic Textual Similarity (STS) task in SemEval-2016, which involve feature selection and Tree-LSTM implementation to generate similarity score.

• Conducted experiments to compare performance between BIO and BILOU encoding schemes for Name Entity Recognition.

• Designed and implemented models for bilingual word representations based on word2vec and achieved state-of-the-art performance.

• Implemented Word2Vec model using Java