Geospatial deep learning
Object detection, instance and semantic segmentation for satellite and drone imagery using Mask R-CNN, DeepLabv3, YOLOv5, and U-Net.
Making Earth more legible.
Turning satellite pixels into planetary insight.
Neha Sharma brings GeoAI research closer to real-world decisions, turning Earth observation into actionable insight for cities, infrastructure, and climate. Her work spans geospatial deep learning, multimodal remote sensing, and digital twins through collaborations with MIT CSAIL and the WCRP Digital Earths effort. She also serves as a GeoAI subject matter expert for Bootcamp in GIS, a California-based GIS edtech company.
- Now GeoAI Developer · MIT CSAIL
- Member WCRP · Digital Earths
- SME Bootcamp in GIS, California-based GIS EdTech
Neha Sharma
Gwalior, India
26.2° N · 78.2° E
- RESEARCHERGeoAI · Earth Observation · Digital Twins · Remote Sensing
- AUTHORCo-editor, Smart Buildings & Cities (T&F, 2025)
- EDUCATORNIT Calicut · CoA GAP · 2,500+ community
- years of industry experience in global R&D centers
- 6+
- papers, chapters, and a co-edited book
- 11
- sessions across NIT Calicut, SPA Bhopal, CoA-TRC & SAIARD/NIUA
- 30+
- people trained through national capacity-building programs
- 200+
- members in her global GeoAI community
- 2,500+
Profile
From designing buildings to decoding the planet.
Trained across architecture, geomatics, and deep learning, Neha builds geospatial AI that is rigorous enough for research and usable enough for planners. She is a founding member of the GeoAI R&D team at Larsen & Toubro, and spent three years at Esri’s GeoAI R&D Center shipping and quality-assuring pretrained models used across ArcGIS.
Today she works remotely from India with MIT CSAIL on a visual data science platform for multimodal knowledge exploration and human–AI collaboration, and contributes to the WCRP Digital Twins – Beyond Physical Systems Working Group.
The data matters. The living world matters more. Her work keeps people, ecosystems, and local knowledge in view as spatial data becomes insight.
01 · Research focus
Where Earth observation meets machine intelligence.
Five connected threads, from the pixel-level models to the Earth system frameworks they ultimately serve.
Multimodal Earth observation
Fusing RGB and SAR for robust building detection and height estimation, using a dataset that spans seven continents.
Digital Earth & digital twins
Connecting human and physical Earth systems for hydroclimate risk, water, agriculture, energy, and urban planning.
Urban air & environmental monitoring
Spatio-temporal analysis of pollution and land use with Sentinel-5P, MODIS, and AURA OMI, including a satellite-based air quality index.
Human–AI collaboration
Visual data science and multimodal knowledge exploration at MIT CSAIL, plus efficient LLM fine-tuning for low-resource settings.
02 · Selected research
From pixels to patterns.
Theses, models, and applied systems across remote sensing, computer vision, GIS, and language models.
Reading rooftops from above
Fine-grained classification of 12 roof types, with a unified DeepLabv3 model that detects buildings and estimates height simultaneously from high-resolution satellite imagery.
Springer paperShip detection in RGB imagery
Led a Mask R-CNN pretrained model for localizing ships in high-resolution imagery, trained on an in-house dataset of 70,000 ships built globally from Maxar imagery.
View modelUrban air quality, in space & time
Mapped pollution hotspots with AURA OMI, MODIS MAIAC, and Sentinel-5P, including the 2020 lockdown, and built a satellite-based spatial air quality index.
View thesisEfficient LLM fine-tuning
Compared LoRA, QLoRA, and prefix-tuning against full fine-tuning on FLAN-T5, cutting GPU memory by 65–70% with minimal accuracy loss.
Full-motion video analytics
Geo-located objects from drone metadata, ran Mask R-CNN detection on video, and shipped a Flask and Leaflet web app for monitoring sites.
Following the fields of Gujarat
District-level cropping-pattern assessment for the 2018 monsoon, examining crop shifts, surplus, and suitability.
View projectToolkit
03 · Publications
Ideas made shareable.
Co-edited book
Smart Buildings and Cities with Remote Sensing and GIS
Co-edited with Dr. Shiv Mohan and Dr. Navneet Munoth, a first-of-its-kind volume bridging remote sensing, GIS, architecture, and planning. Neha also co-authored four chapters in the book. Published with Taylor & Francis (2025), New York, USA.
Discover the bookSatellite Image-Based Spatial Analysis of Urban Air Quality Index
Darshana Rawal, Sindhu Ranganath, Neha Sharma & Anjana Vyas. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLIX-B3-2026, 773–781. ISPRS Congress 2026, Canada.
Fine-Grained Roof Type Classification & Multimodal Multitasking Building Detection & Height Estimation Using High-Resolution Satellite Imagery
First author. International Conference on Data Science and Emerging Technologies (DaSET 2024), Springer, Singapore.
Introduction to Remote Sensing and GIS
In Smart Buildings and Cities with Remote Sensing and GIS, CRC Press · Taylor & Francis.
Land Use Land Cover Mapping of San Diego Using AVIRIS Hyperspectral Imagery
First author. In Smart Buildings and Cities with Remote Sensing and GIS, CRC Press · Taylor & Francis.
More publications (2016–2025)
- 2025 Smart Architectural Site Analysis using Remote Sensing and GIS · book chapter, second author
- 2025 Satellite Missions and Data Products for Architectural and Urban Planning Applications · book chapter, first author
- 2021 GIS-Based Air Quality Index Spatial Model for Indian Cities · Challenges and Solutions for Sustainable Smart City Development, Springer, second author
- 2020 Dramatic & Conspicuous Impact of National Lockdown 2020 on Air Pollution Levels of Ahmedabad District · National Symposium (ISRO, ISRS, ISG), first author
- 2020 CAG Assesses Impact of Lockdown on Air Quality · CRDF Connect magazine, first author
- 2016 Green Building in Hot-Dry Climate – Design Contemplations · Amity University Journal, first author
Latest writing
MEDIUM · 4 NOV 2026
Download ResourceSat LISS-IV (5.8 m) Satellite Imagery from ISRO’s Bhoonidhi Portal
MEDIUM · 24 OCT 2026
How to Download Landsat 8 & Landsat 9 Satellite Imagery from USGS EarthExplorer
ARCGIS BLOG · 15 MAY 2025
Boosting GeoAI Model Performance with Test Time Augmentation
ARCGIS BLOG · 22 DEC 2023
ArcGIS AI Models – Year in Review
The complete record is on ORCID and Google Scholar .
04 · Experience
A path through different scales.
-
Oct 2026 — Present
GeoAI subject matter expert · Bootcamp in GIS California-based GIS EdTech
Contributing GeoAI expertise to a hands-on GIS learning programme.
-
May 2026 — Present
GeoAI Developer · MIT CSAIL Remote from India
Developing geospatial capabilities on a visual data science platform for multimodal knowledge exploration and human–AI collaboration.
-
May 2026 — Present
Member · World Climate Research Programme Geneva
Digital Earths Lighthouse Activity, sponsored by WMO, IOC-UNESCO, and ISC. Advancing Digital Earth and digital twin frameworks that integrate human and physical Earth systems.
-
Feb 2023 — Jan 2026
Product Engineer II / I · Esri GeoAI R&D Center New Delhi
Built and maintained ArcGIS pretrained models on Living Atlas, led the Ship Detection (RGB) model, and ran quality assurance for the
arcgis.learnmodule, AutoML, and time-series tools. Authored documentation, story maps, and blogs. -
Oct 2020 — Feb 2023
Software Engineer, GeoAI · L&T · Mindtree · LTIMindtree Chennai
Founding member of the GeoAI R&D team. Delivered biodiversity big-data pipelines on Azure Databricks for S&P Global, drone video analytics, and deep learning for solar, construction, pool, and land-cover detection.
-
2020
Research Assistant & Teaching Associate · CEPT University Ahmedabad
Contributed to an ISRO-SAC urban air quality project (Sentinel-5P, MODIS) and assisted with the M.Tech Geomatics foundation studio.
05 · Teaching & capacity building
Building skills across disciplines.
Neha has delivered 24 expert and guest sessions across NIT Calicut, SPA Bhopal, CoA-TRC, and SAIARD/NIUA, and has co-organized and taught through the Council of Architecture’s national GAP capacity-building programme.
GeoAI for resilient urban systems
Faculty for the Centre of Excellence in Urban Planning & Design. Delivered 16 hands-on sessions in 2025 and 5 more in June 2026 on GeoAI for disaster management and digital twins for urban managers.
GAP: Bridging Geomatics, Architects & Planners
A national online programme in two levels, from GIS foundations to LULC, UAV, BIM–GIS, and spatial machine learning. Over 200 students and professionals trained.
Guest sessions
- 2025 Google Earth Engine for Spatial Planning · SPA Bhopal
- 2022 Digital Research Methods in Architecture & Planning · CoA-TRC FDP
- 2021 Role of GIS in Smart Transportation · SAIARD / NIUA
06 · Recognition
Milestones along the way.
Awards & distinctions
- 2025Outstanding Achiever AwardMS in Data Science · Liverpool John Moores University
- 2022GICEA Gold MedalM.Tech Geomatics · Gujarat Institute of Civil Engineers and Architects
- 2022MindFest Hackathon WinnerServiceNow Platform Hackathon · Mindtree
- 2021Game Changer Team AwardMindtree NxT Hackathon · GeoAI idea for city cleanliness
- 2020Academic Excellence, Best Capstone & Distinguished StudentM.Tech Geomatics · CEPT University
- 2012Best Student of the Year · MathematicsCarmel Convent School, Gwalior
National exams
- 2022GATE · Geomatics EngineeringAll India Rank 171
- 2020National JRF Eligibility Test (JET)Qualified · ISRO–IIRS, first national edition
- 2018CEPT PG Entrance · GeomaticsAll India Rank 2 · admitted to M.Tech Geomatics at CEPT with a merit scholarship
- 2012–13Joint Entrance Examination (JEE)Cleared Paper 1 & 2 in 2012 and 2013, earning admission to MANIT Bhopal for B.Arch
Professional memberships
07 · Education
Built across disciplines.
MS, Data Science
Liverpool John Moores University
Specialisation in deep learning · DistinctionM.Tech, Geomatics
CEPT University
Remote sensing & GIS · Gold medallistB.Arch, Architecture
MANIT Bhopal
Designing for people & placeRecent certifications
- MIT EECS IAP · GenVizAIMassachusetts Institute of Technology · Jan 2026
- Certificate Program in Generative AIIIT Delhi · Feb–Aug 2025
- PG Diploma in Data Science (Deep Learning)IIIT Bangalore · 2021–22
- Advanced PythonAtmosera · Oct 2024
08 · Journey
A journey mapped across places.
From Gwalior and universities in India to online study at Liverpool John Moores University, remote work with MIT in Boston, and GeoAI teaching for a California GIS edtech team.
A PLACE IN THE JOURNEY
Select a point for the place and mode of collaboration.
A PLACE IN THE JOURNEY
Select a numbered point to reveal its story.
09 · Open source
GitHub contributions.
RESEARCHER PROFILE
@neha-sharma-geoaiLoading public profile…
- Current MIT CompBio
- Former Esri
- Commits
- —
- Issues opened
- —
- Pull requests
- —
- Repositories
- —
- Followers
- —
- Member since
- —
Loading public GitHub metrics…
Counts marked with + reflect publicly visible GitHub activity; private contributions are not included. The daily graph mirrors the public profile calendar through a third-party service (github-contributions-api.jogruber.de).
10 · Collaborate
Let’s map the next question together.
Neha welcomes research collaborations, GeoAI projects, mentoring, and conversations about how spatial data can serve people and the planet.