Leading a group of talented machine learning and backend engineers working on content understanding, LLM and content search.

• VLM reinforcement fine-tuning for engagement prediction

• multi-modal embedding

• semantic id for generative recommendation

• text-to-image retrieval & search ranking.

• search indexing & backend.

Ads optimization.

Landmark Recognition Using 3D Visual Phrase

Used structure from motion technique to get 3D structure from a particular landmark dataset and got 3D virual phrases from space carving. Used these 3D visual phrases to recognize landmarks.

Implemented other image classification methods such as Bag of Words.