About me
I work on auto code generation based on custom/generic prompts for specific use cases in ServiceNow. Before ServiceNow, I worked on the Ranking and personalization stack at Sharechat as a Lead ML Engineer. Personalization and ranking is a multi-objective task involving likes/shares/favs/vplay etc tasks; final ranking is based on these tasks. Much of the experimentation A/B testing, and analysis of the metrics are done as part of the work. Before Sharechat I worked in Defensive Ranking(Responsible AI) techniques at Microsoft as an Applied Scientist. I develop ML algorithms for ranking hosts, URLs and work on sampling and developing models related to the defensive search. Defensive works are related to sensitive topics around BLM, elections, coronavirus, etc. Before Microsoft, I was associated with Walmart where I used to work on neural modeling, used limited labeled data for sequence labeling, and attribute extraction using state-of-the-art NLP techniques. I got my Masters from IIIT Bangalore and completed my thesis under the supervision of , Manish Gupta(Google Research), . I worked on Multi-task question-answering extraction from Video transcripts.
My goal is to teach myself enough in these areas and see how they intersect with the sequential/bandit world. You can find details of my Publications here.
Research and other interests
My area of interest lies in Natural Language Processing, Information Extraction, Information Retrieval, and Reinforcement Learning. I love using these techniques to solve industry-related problems.
Apart from my work, I enjoy traveling and learning about new cultures and languages. I like to find interconnections and variations of languages in different places.
News/Updates
May'22: Our paper titled "Defensive Low Authority Host Predictor" got accepted in Microsoft Journal of Applied Research(MSJAR) 2022.
Nov'20: Our paper titled "Hyper-parameter optimization with REINFORCE and Masked Attention Auto-regressive Density Estimators" got accepted at IEEE Big Data 2020.
Oct'20: Our paper titled "Ultron-AutoML: an open-source, distributed, scalable framework for efficient hyper-parameter optimization" got accepted at IEEE Big Data 2020.
Feb'20: Our paper titled "Learning with Limited Labels via Momentum Damped Differentially Weighted Training" was accepted at 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. Here our focus is on learning from limited labeled data, wherein we aim at jointly leveraging strong supervision data (e.g. explicit judgments) along with weak supervision data (e.g. implicit feedback or labels from the related task) to train neural models.
Dec'19: Our paper titled "Joint Attention Neural Model for Demand Prediction in Online Marketplaces" was accepted at Northern Lights Deep Learning Workshop. Here our focus is on predicting demand for products in online classified marketplace using a joint multi-modal neural model.
Sep'19: Our paper titled "Sequence-aware Reinforcement Learning over Knowledge Graphs" was accepted at REVEAL: A RecSys 2019 Workshop . Here our focus is on explainable recommendation via graphs using RL based approach.
Jul'18: Our paper titled "Neural Attention Reader for Video Comprehension" was accepted at KDD 2018 Deep Learning Day. Here, our focus is on extracting answers for a given question from a pool of videos.