Image Geolocalization
Learning visual representations that connect images with geographic locations.
Research Engineer · India
Hi, I’m Saurabh Patil.
Computer Vision · Visual Geolocalization · Multimodal AI
I am an AI Research Engineer working on visual geolocalization, visual representation learning, and multimodal retrieval. I build systems that connect visual information with geographic context, from model architectures and training pipelines to large-scale retrieval and evaluation.
Research, engineering, and a little time outdoors.
A little about me
About
My research interests include visual representation learning, multimodal learning, and efficient deep learning. I am particularly interested in developing and adapting large-scale vision and language models for challenging real-world tasks. My work also explores image encoders and visual foundation models, with an emphasis on learning robust, transferable representations for recognition, retrieval, and visual understanding. Earlier, at Carnegie Mellon University, I worked on neural operators and machine-learning methods for scientific computing, including data-driven approaches for partial differential equations. I am interested in advancing foundation models and multimodal systems that connect strong visual representations with practical applications across computer vision, intelligent retrieval, and visual intelligence.
I enjoy building practical systems from advances in computer vision, visual representation learning, and multimodal AI.
What I work on
Research interests
A few areas that currently shape my research and engineering interests.
Visual Place Recognition
Robust place retrieval across viewpoint, appearance, and environment changes.
Representation Learning
Learning transferable visual features for retrieval and recognition.
Multimodal AI
Combining vision and language models for richer visual understanding.
Away from the screen
Beyond research
A few things I enjoy outside of work.
Chess
Enjoying strategy, problem-solving, and the occasional competitive game.
Travelling
Discovering new places, cultures, and experiences.
Video games
Enjoying gaming, immersive worlds, and interactive storytelling.
Movies & shows
A fan of good storytelling, films, and TV series.
Fitness
Keeping active through strength training and regular workouts.
Chess.com profile
@saurabhp12
Selected work
Research & projects
A selection of research directions and engineering projects.
Computer Vision · Geospatial AI
Image Geolocalization
Learning visual representations to estimate where a photograph was captured.
View project
Computer Vision · Visual Retrieval
Visual Place Recognition
Recognizing places across viewpoint changes and challenging indoor environments.
View project
Multimodal AI · Retrieval
Multimodal Geolocation and Reranking
Combining visual embeddings and contextual reasoning to improve location candidates.
View project
Trust & Safety · Computer Vision
AI-Generated Image and Deepfake Detection
A planned detection platform for AI-generated imagery and face manipulation.
View projectResearch output
Selected publications
Papers, preprints, and other public research contributions.
Latent Neural PDE Solver: a reduced-order modelling framework for partial differential equations
Zijie Li, Saurabh Patil, Francis Ogoke, Dule Shu, Wilson Zhen, Michael Schneier, John R. Buchanan Jr., and Amir Barati Farimani
A reduced-order modelling framework that learns system dynamics in a mesh-reduced latent space using a nonlinear autoencoder and a temporal prediction model.
CaFA: Global Weather Forecasting with Factorized Attention on Sphere
Zijie Li, Anthony Zhou, Saurabh Patil, and Amir Barati Farimani
A factorized-attention model for global weather forecasting on spherical geometries, designed to reduce the computational cost of attention at global scales.
Hyena Neural Operator for Partial Differential Equations
Saurabh Patil, Zijie Li, and Amir Barati Farimani
Introduces a neural operator based on Hyena's long convolutions and multiplicative gating, and evaluates it on diffusion–reaction and Navier–Stokes equations.
Background
Experience
AI Research Engineer
Graylark Technologies
Oct 2025 – Present
Research on image geolocalization, including patch-level cross-attention between visual and geospatial embeddings, curriculum learning, and retrieval-focused model evaluation.
Research Assistant
IIT Roorkee
Jan 2025 – Sept 2025
Worked on large-scale cross-modal image geolocalization and retrieval, benchmarking, failure-mode analysis, and multimodal reranking using image and location embeddings.
Research Assistant
Carnegie Mellon University
Jan 2023 – Jan 2024
Developed neural-operator and latent-space methods for learning partial differential equation solution operators, including Hyena Neural Operator and the Latent Neural PDE Solver.
Graduate Research Assistant (Intern)
Carnegie Mellon University
May 2022 – Aug 2022
Explored conditional diffusion models for generating steady-state heat-conduction and incompressible-fluid-flow solutions from observational data.
Education
M.S. in Mechanical Engineering · GPA 3.95/4.0 · Carnegie Mellon University · Jan 2021 – Dec 2022
B.Tech. in Mechanical Engineering · GPA 9.13/10 · Vellore Institute of Technology · Jun 2016 – Jun 2020
Get in touch
Have an interesting idea or want to collaborate?
I’m always happy to connect about research, computer vision, and related opportunities.