Staff ML Engineer
2020 - PRESENTServiceNow- Worked on re-platforming Passage AI technologies like intent classification, named entity recognition, machine reading comprehension in the ServiceNow platform.
- Worked within a small team and introduced multilingual capabilities to 15+ languages (from 5) in ServiceNow.
- Built the latest prebuilt models for various machine learning (ML) algorithms.
Technologies: TensorFlow, Keras, Python, JavaLead NLP Engineer
2018 - 2020Passage AI- Designed and built the following systems: named entity recognition, slot filling, knowledgebase relevance and ranking, end-to-end L2 ranker system that improves with user feedback, multi-language capabilities. Was awarded a patent for our work.
- Directly interacted with clients to understand, brainstorm, and design new features to streamline all relevance requirements and was instrumental in customer success in a couple of clients.
- Established coding standards and engineering of our microservices deployed in Kubernetes for low latency and memory footprint.
- Contributed actively to set the NLU direction with CTO. Passage AI was acquired by ServiceNow.
Technologies: Keras, Node.js, PythonSoftware Development Engineer II
2012 - 2020Microsoft- Contributed to the build of the Bing Product Ads' content API from scratch. Worked on all levels of the stack. It currently has around 50 million offers.
- Worked on various components of Bing Knowledge Graph including scaling different parts of the data pipeline and designing and implementing REST APIs to showcase end-to-end capabilities while collaborating with teams in China and India.
- Designed and built a new algorithm to automatically conflate different sources of truth like Wikipedia, IMDB, independent wikis, and knowledgebases. This fixed a significant number of duplicates in our knowledge graph.
- Ideated, designed, and developed patent-pending technology (404123-US-NP) to bring intelligence to office products from auto-filling in MS Word and Excel to fact validation in ms word and outlook. This won the company-wide yearly hackathon in 2016.
Technologies: Pipelines, SQL, C#Research Scientist
2010 - 2020Adometry- Devised, developed, and tested new approaches to improve click fraud detection and ad-analytics products including ad-campaign optimization and ad-attribution analytics.
- Involved in leading/assisting in all stages of the client process (from onboarding to the delivery of results.
- Worked as one of the first research engineers and involved in developing a research strategy. Adometry was later acquired by Google.
Technologies: Java, Python