We are a research-led training studio for the AI era. Darshini teaches GIS, remote sensing and
GeoAI — and the machine learning, automation and cloud workflows that turn Earth observation
into decisions people can act on.
Next cohort starts 5 December 20262 weeks · live online
Participants
273
Institutions
150+
Cohort rating
4.8/5
Peer-reviewed papers
8
Toolkit
Python
Google Earth Engine
InVEST
ArcGIS Pro
QGIS
GeoPandas
XGBoost
PyTorch
SHAP
Sentinel-1/2
Landsat
HEC-RAS
FastAPI
Docker
01 Upcoming workshop · December 2026
Ecological Carbon Storage & Sequestration
How can satellite-derived land-use data tell us what a landscape stores — and what it loses?
Build the whole chain in two weekends: land cover, carbon pools, modelling, mapping and valuation.
Starts5 December 2026
Duration2 weeks · 8 hrs on weekends
FormatLive online · hands-on
IncludesRecorded lectures
What you will learn
InVEST Carbon Model
The core storage and sequestration model, from inputs to outputs.
Google Earth Engine
Cloud-scale land-cover and change detection over your own area.
Land Use / Land Cover
Build and validate LULC layers that the carbon model can trust.
Soil Organic Carbon
Assign credible SOC values by land-cover class and depth.
Above-Ground Biomass
AGB estimation and the assumptions that make or break it.
Below-Ground Biomass
Root carbon, ratios and how to defend your numbers.
Dead Organic Matter
Litter and dead wood — the pool most analyses forget.
Scenarios & Valuation
Zonal summaries, scenario analysis and economic framing.
Skills you leave with
Work with real datasets, not toy examples
Build carbon pools tables from scratch
Run the InVEST model end to end
Create maps and zonal summaries
Analyse scenarios and value the change
Explain limitations and communicate results
02 Flagship programme
GEO-AI Mastery — completed, twice
Our first flagship ran two full parts in 2026, with participants assigned a landslide-susceptible
district each and taught to build a research-grade susceptibility study from raw layers to model.
Completed · Part 1 & Part 260 participants in Part 1 · 4 educators
Every participant was assigned a landslide-susceptible district in India and had to derive,
clean and map their own research-grade conditioning factors — the exact workflow behind our
published Chamoli study.
GeologicalLineament density · Lithology · Soil type, texture & depth · SPI · TRI · TWI
EnvironmentalNDWI · MNDWI · Rainfall
Human activityDistance from roads · Road density · LULC
Part 2
Advanced GeoAI: modelling, deep learning & interpretation
Live weekend classes · recorded sessions for participants
Advanced geospatial data processing, machine learning and deep learning, SHAP-based model
interpretation, and end-to-end GeoAI case studies — taken from the same pipelines the team
publishes with.
XGBoost
Random Forest
MLP-NN
DenseNet
SHAP
Model evaluation
Reproducible notebooks
Taught by Sayantan Mandal, Nilanjan Bal, Gopal Chowdhury
with Arup Baidya · delivered with Revo Map Solutions.
GeoAI workflow · landslide susceptibility
AUC–ROC0.87
Ground control868 GCPs
Conditioning factors18 layers
Model ensembleDenseNet-169 · XGBoost · MLP-NN
ExplainabilitySHAP attribution
Method taught in Part 1 · Journal of the Indian Society of Remote Sensing, 2026
Community poll
Interested in the next GEO-AI cohort?
Tap interested and you'll land straight in our WhatsApp community — batches, dates, free sessions and recordings are announced there first.
150+ learners already in the community · leave anytime
03 Reviews & trust
4.8 / 5, from the people who finished it
Real responses to the GEO-AI Mastery Program Part 1 feedback form — 20 of the 60 participants
replied, and every one of them rated the course 4 or 5 stars.
4.8/5
20 verified responses · Part 1 feedback form
5★16
4★4
3★0
2★0
1★0
100%said they would join the next part (20 of 20)
0ratings below 4 stars
Who was in the room
Students8
Research scholars8
Working professionals4
From Adamas, COEP Pune, University of Delhi, Banaras Hindu University, Symbiosis, Lovely Professional University, Tata Institute of Social Sciences, JNU, University of Kalyani, Samarkand State University and more.
What participants took away
“I really appreciate how approachable the instructors were — they helped every one of us with
our queries. Recreating a whole case study of landslide susceptibility mapping for a district
is very unique and different from other teaching methods. It means we build a portfolio right
out of the classes.”
Preethi Vasudevan · Technical Officer, Wetlands International South Asia · rated 4/5
Every participant left with
A district-scale landslide susceptibility study they built themselves
18 conditioning factors derived, cleaned and mapped
Model outputs with SHAP attribution, not a screenshot
A project they can put in front of a supervisor or an employer
In their words
“
The course was very fruitful. I got to learn a lot about the various data sources and how all the variables play a major role in the occurrence of a disastrous event.
Best course I have ever attended and purchased. The sessions were interactive, all doubts were solved properly, and I got to learn a lot in a short span of time.
Swati DagarStudent · 5/5“
The course was very helpful for basic concept clearance to layer preparations. It was really a hands-on course.
Niloy SarkarResearch Scholar, University of Kalyani · 5/5“
Nothing — everything is perfect.
Somnath SamantaResearch Assistant, PFRI · 5/5“
Thank you for organizing this programme. The first part was especially helpful for beginners.
Dr. Nisha SinghIndependent Researcher, JNU · 4/5“
I really appreciate how approachable the instructors were. Recreating a whole case study of landslide susceptibility mapping for a district is very unique.
Preethi VasudevanTechnical Officer, WISA · 4/5
Use the arrows or swipe. Names appear exactly as participants submitted them.
04 Watch first
See a class before you pay for one
Our webinars are the same material we teach in the paid cohorts — published free on YouTube.
If the teaching style fits, the workshop will too.
Geospatial is our home ground, not our ceiling. The same team now teaches machine learning,
automation and cloud deployment — for geographers, engineers, analysts and anyone who works with data.
01
GIS & Earth Observation
Vector and raster analysis, spatial statistics, Kriging, AHP, ArcGIS Pro and QGIS — plus Sentinel-1/2, Landsat and MODIS processing in ERDAS, ENVI and SNAP.
02
Google Earth Engine
Cloud-scale Earth observation: image collections, cloud masking, compositing, 25+ spectral indices, time series and exports through the JS and Python APIs.
03
GeoAI & Deep Learning
XGBoost, Random Forest, MLP-NN, DenseNet, CNN–LSTM ensembles, GANs for DEM super-resolution, the Segment Anything Model and SHAP for interpretation.
04
AI & Automation
Python automation and ETL, LLM-assisted research and reporting workflows, reproducible notebooks, FastAPI services and model deployment with Docker.
05
Carbon & Climate Analytics
InVEST carbon storage and sequestration, LULC-driven biomass accounting, drought and land-degradation modelling, climate-risk assessment.
Voxel surfaces, digital terrain, CityEngine and scenario modelling — from urban expansion studies to infrastructure risk.
08
Research Methods
Study design, VIF screening, stratified sampling, accuracy assessment, AUC-ROC validation and reproducible research workflows.
Who we train
Students of geography, environmental science, agriculture, engineering and planning
Researchers and Ph.D. scholars building a methods chapter
Analysts and professionals moving into AI, GIS or data science
Institutions and teams needing applied, hands-on capacity building
How we deliver
Weekend cohorts with recorded lectures
1-day workshops and 3–5-day bootcamps
2–4 week certificate programmes
Customised institutional training and research mentoring
06 Research
Published, not promised
Since 2022 the team has authored eight peer-reviewed journal papers — the majority in Scopus-indexed
Q1/Q2 journals. What we teach is what we publish.
Featured study · the method taught in Part 1
A Novel Deep Learning Approach for GIS-Based Landslide Susceptibility Prediction in the Himalayas
Mandal, S. & Saha, A.K. (2026) · Journal of the Indian Society of Remote Sensing
89.46%Accuracy
0.87AUC–ROC
868Field GCPs
18Factors
A custom DenseNet-169 architecture benchmarked against MLP-NN and XGBoost across the 7,951 km² Chamoli district, with SHAP identifying geomorphology, land use, road proximity, slope and soil type as the leading predictors.
Currently under review: “Decoding the Non-Farm Transformation: a t-SNE–SHAP Stacking ML Audit” and “Assessing Multidimensional Sustainability in Rural Non-Farm Transformation”.
07 Impact & community
273 participants. 150+ institutions
Three webinars and one flagship programme, across India and abroad — with participants from
Ethiopia, Poland, Pakistan, Bangladesh and Uzbekistan.
273
Registrations across 3 webinars & the flagship
249
Unique participants
150+
Partner academic & professional institutions
4.8/5
Average cohort rating
100%
Would join the next programme
Represented institutions include
Banaras Hindu University
Vidyasagar University
University of North Bengal
University of Delhi
University of Allahabad
Adamas University
University of Calcutta
Savitribai Phule Pune University
West Bengal State University
The University of Burdwan
Symbiosis Institute of Geoinformatics
Central University of South Bihar
Andhra University
Jawaharlal Nehru University
IIT Guwahati
IIT Kharagpur
COEP Technological University
Tata Institute of Social Sciences
IIM Kozhikode
Jadavpur University
Kazi Nazrul University
Central University of Punjab
Samarkand State University
University of Rajshahi
FTVTI Addis Ababa
WhatsApp community
150+ learners. One group. Every announcement first.
New batches, free webinars, recorded lectures, job leads and doubt threads — share your interests and we will tell you what is coming next, whether that is carbon, GeoAI or automation.
A remote-first team of doctoral researchers at the University of Delhi and working GeoAI engineers,
communicating from New Delhi — so the material is never far from peer review or production.
SM
Sayantan Mandal
Founder & Lead — GeoAI & Machine Learning Engineer
Leads technical and research direction. Four-plus years translating deep-learning research into
production geospatial systems; currently building AQWeather, a farm-intelligence platform at
AquaCarta, alongside doctoral research on AI/ML multi-hazard assessment in the Himalayas.
GeoAI
Deep learning
MLOps
LLM pipelines
GEE
Ph.D. Geography, Delhi School of Economics, University of Delhi (expected Nov 2026)
Developed the CRPAG-DSHAT GAN for Himalayan DEM repair
Early Segment Anything Model application to building exposure
ANRF International Travel Grant 2025 · Best Paper, INCA 2023
A UGC Senior Research Fellow and Ph.D. scholar in Geography at the University of Delhi,
specialising in geo-environmental hazards, climate-resilient urban and peri-urban planning,
geocomputation and digital-twin modelling.
Geostatistics
Simulation
Digital twins
XGBoost
3D modelling
Ph.D. Geography, Delhi School of Economics (since Feb 2024)
M.A. Geography, Banaras Hindu University — CGPA 8.48/10
Qualified WBCS-SET Assistant Professor (2022) and UGC-NET JRF
Published in Annals of the AAG and The Professional Geographer
Leads the classroom. A UGC Senior Research Fellow and Ph.D. scholar in Geoinformatics at the
University of Delhi, with five peer-reviewed publications on agricultural distress, drought
zonation and land degradation using GIS, Google Earth Engine and hybrid deep-learning models.
GIS / RS
GEE
Hybrid deep ensembles
Python
SHAP
Ph.D. Geography, Delhi School of Economics (ongoing)
M.A. Geography, Visva-Bharati University — 81%, A++
Prof. N.P. Aiyyar Young Geographers Award, 47th Indian Geography Congress
First Prize, Best Paper, NAGI 7th International Conference (2024)
Revo Map Solutions“Mapping Your Success” — co-hosts the Breaking into GeoAI series and co-delivered the GEO-AI Mastery Program
Department of Geography, Delhi School of Economics, University of DelhiThe academic research base behind the team's publications, under Prof. Ashis Kumar Saha
Tata Institute of Social Sciences, MumbaiSchool of Disaster Management — collaborative geospatial hazard research
09 Events
What is coming, and what already happened
One live workshop is open for registration. Past webinars are on YouTube — watch them before you
decide anything.
5 Dec2026
Ecological Carbon Storage & Sequestration Open
2 weeks · 8 hours on weekends · live online · recorded
InVEST carbon model, Google Earth Engine, carbon pools, scenario analysis and valuation — from LULC to a decision-ready summary.
20-hour Part 1 · advanced Part 2 · 60 participants
Landslide pre-conditioning factors to deep-learning susceptibility modelling, with SHAP interpretation and end-to-end projects.
14 Mar2026
Breaking into GeoAI
With Revo Map Solutions · free webinar
GeoAI trends, practical ML case studies and building an end-to-end hazard pipeline from landslide to hydrodynamic simulation.
29 Mar2026
XGBoost for Crop Yield Estimation
Hands-on notebook session
Gradient-boosted modelling for crop yield, from feature construction to predictions and reproducible notebook workflows.
12 Apr2026
Hydro-Spatial Groundwater Blueprint
AI-based groundwater potential zonation
13 conditioning factors, AHP versus ML versus deep learning, VIF screening and independent well validation (DLNN AUC 0.89).
10 Questions
Before you register
Still unsure? Join the community and ask — a founder answers, not a call centre.
What exactly is the December workshop?
Ecological Carbon Storage & Sequestration — a two-week, hands-on online workshop (8 hours across weekends) starting 5 December 2026. You will build LULC layers in Google Earth Engine, assemble carbon pools (AGB, BGB, SOC, dead organic matter), run the InVEST carbon model, map storage and sequestration, and finish with scenario analysis and economic valuation.
What does it cost, and are there discounts?
₹1,499 instead of ₹2,499 — a 40% discount. If three or more people register together the fee is ₹999 per person. Registration is through the Google Form on this page; we will share payment details after your form is received.
I am not from a geospatial background. Can I join?
Yes. The workshop starts from land-cover fundamentals and uses point-and-click InVEST plus guided Earth Engine scripts, so no prior Python is required. If you are comfortable with basic file handling, you are ready. Carbon, ecology, agriculture, economics and policy backgrounds are all welcome.
What happened to the GEO-AI Mastery Program?
It is complete — Part 1 and Part 2 both ran in 2026 with 60 participants, rated 4.8/5, and every respondent said they would join the next part. The next cohort is not dated yet. Tap “I'm interested” in the programme section to join the WhatsApp community and you will hear about it first.
Do you only teach geospatial topics?
No. Geospatial is where we are strongest, but we are expanding into AI and automation — Python automation, LLM-assisted workflows, model deployment and MLOps. Tell us in the community what you want next; the calendar follows demand.
Do you train institutions and teams?
Yes — 1-day workshops, 3–5-day bootcamps, 2–4 week certificate programmes and customised corporate or university training, including research mentoring for departments and labs.
Where are you based?
We are a remote-first, virtual team communicating from New Delhi, India. All programmes are delivered live online, so you can join from anywhere.
Is the money refundable?
If you tell us before the first session that it is not right for you, we will refund the fee. After the first session, we will transfer the amount to a later cohort instead. Write to us — we would rather fix the fit than keep a fee.
11 Contact
Let's talk — wherever you are
Registering is a form. Everything else — group discounts, institutional training, research
collaboration — starts with a message.
