I’m a Senior Applied Scientist in the Generative AI space at Microsoft. I own the Smart-Reply Microsoft Research (MSR) project, that saves ~294,000 hours/year by predicting accurate solutions to human queries leading to ~$3.6 Million cost savings/year. Currently, I work on long-tailed domain-specific Conversational AI LLMs in the Copilot project (press, video). I have a Masters in Computer Science from Columbia University and a Bachelors in Computer Science from IIT Guwahati.
Few months after graduating from Columbia University in 2019, I realized my desire to pursue a PhD and enter academia. I started working with Professors Zhou Yu, Smaranda Muresan, and Ching-Yung Lin after work-hours to create a strong candidature for a good PhD program. I did independent research and published at LREC, NeurIPS, and ACM Multimedia. Now, I am working on first-author long papers in *CL main conferences after work-hours. I am an advisory board member of SNF NGO (letter, approval) working on AI-based education of underprivileged children, that has achieved remarkable results.
2024: [Mar] Fourth promotion in three years at Microsoft as a Scientist
2024: [Feb] General Availability Release of Dyanmics 365 Offering, as AI Lead
2023: [Jan] Third promotion in two years at Microsoft as a Scientist
2023: [Jan] NLP Paper accepted at WWW’23 (acceptance rate 19.8%)
2022: Two invited talks delivered at Columbia University, and NIST
2022: [Sep] Ranked #1 in NIST (US Gov.) TRECVid on Multimodal Language Understanding (DVU)
2022: [Jul] Invited Talk at Columbia University (PDL - Engineering) on becoming an industry researcher
2022: Formal Mentor for a Microsoft Research (MSR) JEM NLP PhD Summer Intern
2021: Invited talk at NIST for my work on long form datsets - multi-modal semantic inference
2021: Ranked #1 in ACM Grand Challenge on Multimodal Semantic Understanding in long-form datasets
2020: Three peer-reviewed research papers published with external collaboration
2019: Honored as Class Marshal: Columbia Class of ‘19 for “demonstrating achievement in academics”.
2018: Featured in Uber AI's blog for going above and beyond.
2015: Hour-long session with Director of Research, Google : Peter Norvig; “Limitations of Deep Learning”
Hobbies: I am a Violinist, Pianist, Portraitist, run Half-Marathons, dance Salsa, and solve Rubik’s Cube.
Discipline: Workout 4 days/week since past two years; hold orange-belt in Shito-Ryu-Genbu-Kai Karate.
Banking: Was a macro delta wall-street banker earlier - with each trade ranging between $5MM-$50MM.
Highlight Paper:
Highlight Demo:
Contact:
