Hi, I am Shasvat Desai
Hi, I am Shasvat Desai
Embedded Files
Staff Applied Scientist, Sponsored Search, Walmart
Staff Applied Scientist, Sponsored Search, Walmart
Turning pixels and letters into insights, one gradient descent at a time 😀
Turning pixels and letters into insights, one gradient descent at a time 😀
ABOUT ME
ABOUT ME
I am currently working as a Staff Scientist at Walmart, where I am involved in developing algorithms for sponsored search. I am the technical Lead in AdTech Science focused on semantic query understanding, product retrieval, and ranking for sponsored search in e-commerce. I lead end-to-end development of ML and LLM systems—from data strategy and model training to distillation, low-latency serving, and online experimentation. My work connects deep semantic understanding with large-scale retrieval, ranking, and optimization to improve relevance, engagement, and revenue. I partner closely with Engineering, Product, Analytics, and Business teams to drive measurable impact through rigorous evaluation and A/B testing.
I am currently working as a Staff Scientist at Walmart, where I am involved in developing algorithms for sponsored search. I am the technical Lead in AdTech Science focused on semantic query understanding, product retrieval, and ranking for sponsored search in e-commerce. I lead end-to-end development of ML and LLM systems—from data strategy and model training to distillation, low-latency serving, and online experimentation. My work connects deep semantic understanding with large-scale retrieval, ranking, and optimization to improve relevance, engagement, and revenue. I partner closely with Engineering, Product, Analytics, and Business teams to drive measurable impact through rigorous evaluation and A/B testing.
Previsouly, I worked as a Senior Computer Vision Scientist at the Intelligent Retail Lab (IRL), Walmart, where I was involved in developing algorithms to better understand what’s happening inside the store through an array of sensors, cameras, and processors to reduce retail shrinkage at self-checkout systems (SCO). My work primarily lied in the domain of scene understanding utilizing spatio-temporal scene graphs, multi-view geometry (epipolar and projective), object detection and tracking, and multi-modal retrieval integrating visual, textual, and product catalog data for precise SKU item recognition across Walmart’s extensive catalog.
Previsouly, I worked as a Senior Computer Vision Scientist at the Intelligent Retail Lab (IRL), Walmart, where I was involved in developing algorithms to better understand what’s happening inside the store through an array of sensors, cameras, and processors to reduce retail shrinkage at self-checkout systems (SCO). My work primarily lied in the domain of scene understanding utilizing spatio-temporal scene graphs, multi-view geometry (epipolar and projective), object detection and tracking, and multi-modal retrieval integrating visual, textual, and product catalog data for precise SKU item recognition across Walmart’s extensive catalog.
Before that, I was a Computer Vision Scientist at Orbital Insight, where I worked with satellite, aerial and drone images in the geospatail domain. My work spanned across multiple domains ranging from object detection of military and naval assets, landuse segmentation and monitoring (deforestation, agricultural expansion, infeastructure development) in multispectral imagery to domain-adaptation, super-resolution, and semi-supervised learning.
Before that, I was a Computer Vision Scientist at Orbital Insight, where I worked with satellite, aerial and drone images in the geospatail domain. My work spanned across multiple domains ranging from object detection of military and naval assets, landuse segmentation and monitoring (deforestation, agricultural expansion, infeastructure development) in multispectral imagery to domain-adaptation, super-resolution, and semi-supervised learning.
Before working at Orbital Insight, I was a Masters' student at the University of Massachusetts, Amherst, where I specialized in Machine Learning and Computer Vision. During my time here, I created the one of its kind multimodal salient pedestrian dataset with pixel level-annotation which has since been adopted widely in the automotive and multimodal research domain.
Before working at Orbital Insight, I was a Masters' student at the University of Massachusetts, Amherst, where I specialized in Machine Learning and Computer Vision. During my time here, I created the one of its kind multimodal salient pedestrian dataset with pixel level-annotation which has since been adopted widely in the automotive and multimodal research domain.
In my free time, I enjoy hiking, photography and playing board games!
In my free time, I enjoy hiking, photography and playing board games!
News
News
June 2026: Gave a virtual talk at the Toronto Machine Learning Summit on my work on user intent understanding for search systems.
June 2026: Gave a virtual talk at the Toronto Machine Learning Summit on my work on user intent understanding for search systems.
May 2026: Five papers accepted to ACM Special Interest Group on Information Retreival, 2026 (SIGIR 2026).
May 2026: Five papers accepted to ACM Special Interest Group on Information Retreival, 2026 (SIGIR 2026).
(Industry Track) Unified Supervision for Walmart's Sponsored Search Retrieval via Joint Semantic Relevance and Behavioral Engagement Modeling
(eCommerce Workshop) INSPIRE: Intent-aware Neural Sponsored Product Retrieval for E-commerce
(eCommerce Workshop) AutoRelAnnotator: Calibrated Model Cascades for Cost-Efficient Relevance Evaluation in Sponsored Search
(eCommerce Workshop) Scaling Dense Retrieval with LLM-Annotated Training Data: Structured Mining and Progressive Curriculum for E-Commerce Sponsored Search
(eCommerce Workshop) Unified Multi-Task Relevance Modeling for E-Commerce: Comparing Task Routing Architectures Across LLMs and Cross-Encoders
May 2026: Served as a panelist at the Data Infrastructure Summit organized by the AI Accelerator Institute talking about "Build, buy, or borrow: How engineering teams are making data stack decisions that don't haunt them at renewal time"
May 2026: Served as a panelist at the Data Infrastructure Summit organized by the AI Accelerator Institute talking about "Build, buy, or borrow: How engineering teams are making data stack decisions that don't haunt them at renewal time"
June 2025: Patent granted for my prior work at Orbital Insight on Fine-Grained Aircraft Classification from Aerial Imagery.
June 2025: Patent granted for my prior work at Orbital Insight on Fine-Grained Aircraft Classification from Aerial Imagery.
May 2025: Recognized with Bravo Award at Walmart for establishing the ground work in the domain of intent understanding for sponsored search.
May 2025: Recognized with Bravo Award at Walmart for establishing the ground work in the domain of intent understanding for sponsored search.
February 2025: Started a Staff Scientist position in Walmart's Sponsored Search team.
February 2025: Started a Staff Scientist position in Walmart's Sponsored Search team.
October 2024: Recognized with Impact Award at Walmart. This award reflects a commitment to innovation, collaboration and exemplifying values while driving meaningful business impact on fundamental work on temporal aggregation of multiple signals for real-time decision making.
October 2024: Recognized with Impact Award at Walmart. This award reflects a commitment to innovation, collaboration and exemplifying values while driving meaningful business impact on fundamental work on temporal aggregation of multiple signals for real-time decision making.
September 2024: New Medium blogposts on Paradigms on Object Detection and Object Tracking with DeepSORT.
September 2024: New Medium blogposts on Paradigms on Object Detection and Object Tracking with DeepSORT.
July 2024: My team's work on developing scene understanding algorithm utilizing spatio-temporal scene graph representation learning to model interactions and identify shopping behaviors and patterns was released to Walmart customers, making the checkout experience even more smoother.
July 2024: My team's work on developing scene understanding algorithm utilizing spatio-temporal scene graph representation learning to model interactions and identify shopping behaviors and patterns was released to Walmart customers, making the checkout experience even more smoother.
April 2023: My work on using synthetic data for pre-training object detection models got accepted for presentation at Walmart AI Summit Conference 2023. I demonstrated this work across the board to engineers, data scientists, and executives, showcasing the impact of synthetic data on improving detection accuracy and efficiency, as well as its potential for scaling AI solutions across Walmart's vast retail ecosystem. This presentation highlighted the technical depth and strategic value of leveraging synthetic pre-training for large-scale retail applications.
April 2023: My work on using synthetic data for pre-training object detection models got accepted for presentation at Walmart AI Summit Conference 2023. I demonstrated this work across the board to engineers, data scientists, and executives, showcasing the impact of synthetic data on improving detection accuracy and efficiency, as well as its potential for scaling AI solutions across Walmart's vast retail ecosystem. This presentation highlighted the technical depth and strategic value of leveraging synthetic pre-training for large-scale retail applications.
May 2022: Started working as a Senior Computer Vision Scientist at the Intelligent Retail Lab, Walmart, Hoboken, NJ
May 2022: Started working as a Senior Computer Vision Scientist at the Intelligent Retail Lab, Walmart, Hoboken, NJ
October 2021: Our paper titled Active Learning for Improved Semi-Supervised Semantic Segmentation in Satellite Images got accepted to the IEEE Winter Conference on Applications of Computer Vision 2022, to be held at Waikoloa, Hawaii.
October 2021: Our paper titled Active Learning for Improved Semi-Supervised Semantic Segmentation in Satellite Images got accepted to the IEEE Winter Conference on Applications of Computer Vision 2022, to be held at Waikoloa, Hawaii.
Spring 2019: Graduated with a Masters in Computer Science from the University of Massachusetts, Amherst
Spring 2019: Graduated with a Masters in Computer Science from the University of Massachusetts, Amherst
April 2019: Our paper Pedestrian Detection in Thermal Images using Saliency Maps accepted to the IEEE Workshop on Perception Beyond the Visible Spectrum at CVPR, 2019 (Long Beach, California)
April 2019: Our paper Pedestrian Detection in Thermal Images using Saliency Maps accepted to the IEEE Workshop on Perception Beyond the Visible Spectrum at CVPR, 2019 (Long Beach, California)
Summer 2018: Received the DAAD RISE Professional Fellowship to pursue an internship at Bayer, Leverkusen, Germany
Summer 2018: Received the DAAD RISE Professional Fellowship to pursue an internship at Bayer, Leverkusen, Germany
Fall 2017: Started a Masters program in Computer Science at the University of Massachusetts, Amherst, USA
Fall 2017: Started a Masters program in Computer Science at the University of Massachusetts, Amherst, USA
Spring 2017: Graduated from Veermata Jijabai Technological Institute (VJTI), Mumbai, India with a Bachelors in Computer Engineering
Spring 2017: Graduated from Veermata Jijabai Technological Institute (VJTI), Mumbai, India with a Bachelors in Computer Engineering
Get in touch at shasvat.desai@gmail. com
Get in touch at shasvat.desai@gmail. com
Page updated
Google Sites
Report abuse