Senior Data Scientist · Microsoft

Building intelligent systems at worldwide scale.

Data scientist and engineer with 10+ years across commerce, Bing Travel, product intelligence, and knowledge graphs. I turn difficult machine-learning problems into reliable products with measurable business impact.

Explore my work Start a conversation
10+years
building
75MBing Travel users
5M+Copilot+ PCs
3Bpages mined
Hyderabad, India Microsoft Commerce Core · May 2025—Present

01 / IMPACT

Numbers with a story behind them.

1K

Sellers supported

Improved quote-creation workflows through decision support systems and agentic automation in Commerce Core.

Decision supportFaster quoting
Global reach 75M

Bing Travel users

Led a four-person team creating travel experiences worldwide.

Product engagement 1.4M

Daily active users

Reached 1.4M DAU within 1.5 years of Bing Travel's 2021 launch.

Search quality +20%

Hotel price coverage

Improved query processing and satisfaction across five million hotels.

20%

02 / EXPERIENCE

A decade of shipping
ambitious systems.

From research prototypes to high-availability services and consumer experiences at global scale.

2025—NOWMicrosoft

Commerce Core

Senior Data Scientist

DECISION SCIENCE
  • Spearhead data science initiatives focused on measurable business impact through evaluation frameworks and experimentation.
  • Develop decision support systems in Commerce Core, enhancing agentic automation and quote-creation processes.
  • Improved quoting workflows for 1,000 sellers.
ExperimentationEvaluationDecision supportAgentic automation
2020—2025Microsoft

Bing Travel

Senior Data Scientist

SEARCH + ML
  • Led a four-member team building Bing Travel experiences for 75 million worldwide users, reaching 1.4 million daily active users.
  • Created immersive destination experiences including diverse images, panoramas, and flight-simulator videos for 20,000+ destinations; earned a 4.2/5 “Wow” feature score.
  • Improved hotel search precision and relevance, increasing price coverage and satisfaction by 20% across five million hotels and reaching neutral quality in worldwide Google-comparison analysis.
  • Partnered across Microsoft to promote Travel experiences on Edge and Windows.
Information retrievalRankingComputer visionTeam leadership
2019Avalara

HS Code Classification

Software Engineer · Machine Learning

ML SYSTEMS

Built a Harmonized Tariff Code classification engine for cross-border products used by clients including Amazon US. Turned sparse catalog and webpage signals into models for high-volume tariff chapters, automating the workflow with scikit-learn and MLflow as a classification service.

scikit-learnMLflowClassificationProduct data
2017—2019Indix · Avalara

World's Largest Product Catalog

Software Engineer

DATA PLATFORM

Developed a highly available, event-sourced product metadata service for taxonomy, categories, and brands that backed Samsung Mall. Built deduplication, MPN normalization, and identifier enrichment components mining three billion crawled webpages.

Event sourcingData pipelinesMetadataHigh availability
2015—2017Microsoft

Knowledge Graph for Bing · Satori

Software Engineer

KNOWLEDGE GRAPH
  • Created automated lexical and semantic schema mapping to onboard structured and semi-structured sources into Bing's knowledge graph.
  • Built ingestion and source-configuration workflows on Microsoft's Cosmos big-data platform.
  • Managed complete ingestion for World Bank, CIA, Census, and Wikia data sources.
Knowledge graphsSchema mappingBig dataSemantic systems

03 / RESEARCH

Machine learning,
from first principles.

9.3/ 10 CGPA
IIT Kanpur
MASTER'S THESIS2014—2015

Dr. H. Karnick · IIT Kanpur
Dr. P. Jain · Microsoft Research India

Fast Approximate
Composite Machines

Designed FACM, a hierarchical data-partitioning method that scales Kernel SVM training and prediction on large real-world datasets while retaining accuracy close to RBF-SVM.

Faster training and prediction than LDKL on most benchmark datasets
Validated on complex nonlinear binary classification
Benchmarked on Forest, MNIST, and other large datasets
AALTO UNIVERSITY · FINLAND2014

Refinement of predicted contact maps

Under Prof. Erik Aurell, developed a generalized LIBSVM-based Java method for protein structure predictions. Enhanced predictions in 25% of proteins and improved the existing probabilistic model overall.

IBM RESEARCH LAB · INDIA2013

AMPLE · Adaptive Personalized Learning

Under Dr. Sumit Negi, developed a Moodle-integrated system that categorized content by comprehension burden and recommended learning material using each learner's skill profile.

04 / EDUCATION + RECOGNITION

A strong technical foundation.

IITK
2010—2015

B.Tech. + M.Tech.
Dual Degree in CSE

Indian Institute of Technology Kanpur

9.3/10
302All India Rank
IIT-JEE 2010
166All India Rank
AIEEE 2010
KVPY Fellowship
IISc Bangalore
Academic Excellence
Top 5% · twice

05 / PROJECTS + COMMUNITY

Curiosity beyond
the day job.

Compilers, operating systems, recommendation engines, open source, and mentorship.

01 · LANGUAGE SYSTEMS

Dynamic binding in Eiffel

Extended a C++ Eiffel compiler with virtual tables and dynamic polymorphism after studying late binding in C++, Java, and SmallTalk.

C++ · Lex/Yacc · Compilers
02 · RECOMMENDATION

Movie rating prediction

Combined ratings and demographic correlations in collaborative filtering, outperforming existing methods on MovieLens.

Matlab · Weka · Java
03 · SYSTEMS

Pintos + 24-puzzle AI

Implemented threading, virtual memory, and filesystem primitives in Pintos; created ANN-based heuristic learning for the 24-puzzle.

OS internals · ANNs
04 · PRODUCT BUILDING

Laboratory data system

Built a data management system with a WYSIWYG editor and voice input for practical laboratory workflows.

HTML5 · PHP · Bootstrap

IDC Pinnacle AwardAI Usage in Microsoft products · Hackathon 2023

Open-source contributorMicrosoft Fluid, now Microsoft Loop

Academic mentorIIT Kanpur Counselling Service · 2011—2012

Worldwide HackathonTravel stories idea · 2020 and 2022