"We are the first generation to feel the impact of climate change and the last generation that can do something about it." - Barack Obama
A tool that helps users visualize Climate Geospatial data using Generative AI.
Team Role
Worked as Product Manager and Product Designer with a Team of 2 people:
Dhruv Vairagi (PD and PM)
Aditi Bhat (Software Engineer)
What I did
Created the Product Vision Created UX Research Study
Designed the UX and UI
Usability & Accessibility Study
Strategized AI PRD
Created KPIs & Eval Benchmarks
Created Roadmap & GTM
Duration
1 month
Platform
Desktop Application
Background Research
Climate risk and natural disaster occurrence is increasing every year with rising temperatures and changing predictions. Where we invest our money in real estate and how we design our architecture sites is heavily informed by these risk factors.
Why do Urban Planners and Real Estate Advisors need geospatial climate data?
Architects, Real Estate Advisors and Urban Designers today are using newer technologies apart from collaborating with sustainable consultants to find climate risk factors for the sites that they are designing. I sought to explore the need for an application that would make access to these risk data easier for everyone to make informed decisions.
User Interviews and Surveys
The research kicked off by a background research and understanding what existing users in the AEC industry feel about climate risk data and site analysis. This was done through a survey of 51 people and interviews with 6 users.
Users do not have easy access to geospatial climate data
Users rely on sustainable consultants and GIS analysts
Users do not include climate data in designing developments
Users don’t know which data to look for.
Affinity Mapping
Then the issues, ideas and motivations were organized by major themes through affinity mapping.
Challenges and Insights
Based on the current problems listed by the users
1. Climate Change Awareness
Most property owners, architects and even some real estate advisors are not aware of climate change patterns and what risks their sites face as per the IPCC SSP scenarios.
Therefore, it is necessary to educate customers where we are headed.
2. Easy access to Climate data
Most of the times, climate GIS data is needed to assess a site with geospatial data that is not open source or is not easily accessible. Sometimes users do not know where to look for each data separately for different climate issues as it isn't consolidated.
3. Nature Based Solutions
Research has proven that Nature based solutions or NBS has can help in climate risk mitigation and adaptation. Hence their economic benefits need to be quantified so that urban designers and architects can use them and convince developers.
AI product strategy
Develop an easy to use, responsive website for users in the AEC industry to visualize climate risk maps and provide economic benefits of Nature based solutions for their sites using Gen AI while reducing complex steps to access latest climate risk data.
Product Vision and Objectives
Vision:
To empower AEC professionals to use climate risk data for decision making
To help access site and climate data maps
AI-powered
Geospatial Maps and NBS
AI-powered
economic benefit analysis
User Personas
After creating the product goal and vision, customer personas were created who would be primarily using the platform
ALAN (Architect)
Job: Architect at Gensler, NYC
Age: 23 years old
Tools: Revit, CAD, Adobe Creative Suite
Goal: Create LEED certified climate responsive buildings.
Pain Points: Don’t have access to climate geospatial projections.
KATE (Urban Planner)
Job: Urban Planner at AECOM, Dallas
Age: 28 years old
Tools: ArcGIS pro, Adobe Creative Suite, MS Office
Goal: Create Climate Action Plans for cities and state futures.
Pain Points: Have to dig deep and create geo projections
ADITI (Real Estate Advisor)
Job: Real Estate Advisor at HR&A, Chicago
Age: 31 years old
Tools: Tableau, R, PowerBI, MS Office, Presentation
Goal: Maximize profits & do market and site projections for development
Pain Points: Have to pay consultants a lot and don’t get quick access to data.
Competitive Research for Climate Data Providers
A competitor audit was done for existing climate data providers which most of the users use
Reask: Reask uses AI to predict weather risks such as Cyclone risks for various industries
Risk Factor assesses the risk of environmental threats for individual properties across the United States
Climate Central's Coastal Risk Screening Tool helps visualize coastal flood GIS data for next 100 years
The competitor audit resulted in the analysis that there is data available, but not catered towards the AEC professionals to use and download, and that its segregated heavily.
Competitive Research for AI powered Geospatial LLMs
A competitor audit was done for various geospatial LLMs too as that we can understand how users get geospatial data currently using AI and GPTs
Text to Map / Text to CSV / Geospatial AI
MapsGPT
AINO
ChatGPT
Competitor Analysis
Minimum Viable Product (MVP) - GeoPlan
An AI-powered Climate GIS dashboard that provides future climate maps and solutions to planners.
Features & RICE Scores* for Prioritization:
AI Chatbot: where users can input queries on what to visualize & its solutions. (20)
Data Integration:
with Global Climate Models for History & Real Time Data. (10)
AI Powered Visualizations:
Predictive climate modeling & generating dynamic maps. (50)
AI suggested Nature-Based-Solutions for climate issues. (20)
AI-powered cost-benefit analysis for quantifying benefits context specific NBS. (30)
RICE Scores were calculated to understand feature prioritization
*RICE scores are calculated as (Reach * Impact * Confidence) / Effort
AI Chatbot
Economic Benefits of NBS
Global Climate Models
Low Value High
AI Powered Visualizations
AI based NBS
Low Effort High
Product Roadmap
A product roadmap was created to ensure a timeline of the process and delegate tasks
Q1 2024: Completion of requirement gathering and initial design phase.
Q2 2024: Development of AI models and integration of data sources.
Q3 2024: First round of user testing and feedback.
Q4 2024: Official launch and continuous monitoring for improvements.
Future Dev
- Release New Updates
- Expansion in New Market
- User feedback & Iterate
KPIs
Circular Economy Business Model Canvas
Resource Loop
Key Partners
Govt Agencies
Universities
Real Estate Developers
Planners
Policy Advisors
Consultants
Managers
Key Activities
Climate Data collection and analysis
NBS Solutions
Marketing
Key Resources
Analytics infrastructure
NBS & Design expertise
Online Forum
Marketing budget
Value Proposition
- Location-specific climate data and analytics
- GIS data for 2100 years
- Nature Based Solutions
- A learning Community
- AI Chatbot and Predictions
Customer Relationships
Training
Regular updates
Community forums
Channels
Online platform
Partnerships
Digital marketing
Customer Segments
Govt Agencies
Universities
Real Estate Developers
Planners
Policy Advisors
Consultants
Managers
Cost Structure
Data infrastructure and maintenance costs
Research and development expenses
Marketing and outreach expenses
Personnel and overhead costs
Revenue
Subscription-based model for access to premium data
Commission-based Climate insurance
API valuation & data integration
KPIs to measure the success
GOAL
🎯 Provide AI-powered visualizations and analytics for Climate Change Predictions
🎯 AI-powered solutions for climate change built-environment issues.
🎯 Intuitive UX/UI for better collaboration
🎯AI Chatbot UX
🎯 Centralize access to diverse climate data
METRIC
#️⃣ Data Accuracy mAP (mean average precision)
#️⃣ Data Accuracy, Results Mapping
NPS (Net promoter score)
#️⃣ Retention, ROI, NPS
#️⃣ WER & Goal Completion
#️⃣ Data repository
QUESTION
❓How accurate and true is the data visualization?
❓How accurate are the solutions? And what are the metrics?
❓How easy is it for the user to interact?
❓How easy is it to decode the user’s prompts?
❓Which data is missing and not valid?
Entry Point(s)
Major Step(s)
Sub-step(s)
Data Flow
Generated Output
AI Chatbot User Flow
Users Login to GeoPlan
Users add site data and ask climate data
AI Model analyzes query using NLP
Model extracts data from geospatial APIs.
Model generates NB solutions
The user flow shows how the AI chatbot can help in the human-in-loop user flow and automate tasks with the help of NLP
Information Architecture
Before jumping into wireframing and prototyping, an information architecture or a site map was developed, so that we can understand how the site layout would function.
GeoPlan AI
UX Prototype
Onboarding
The user signs up and creates an account after which the UI of GeoPlan is a maps interface which allows the user to create a site boundary for their site and ask the LLM some pre-generated questions or ask it general climate information.
AI Chatbot
The chatbot allows the user to search for climate geospatial data for current and future scenarios. This is accelerated and augmented with the help for Natural Language Processing which helps the user’s response to be decoded and understood by the LLM and generate the relevant data.
AI Powered GIS Climate Data
The climate geospatial data for future predictions is pre-generated with the help of global climate models such as the IPCC SSPs and stored in cloud storage. The LLM will extract the data and present it to the user to be able to download and work with it.
AI-Powered Nature Based Solutions Quantification
The LLM is also trained to quantify the economic benefits of a NBS to get relevant information of the site such as improvements in Land Surface temperatures, property costs, public health and tourism.
Model & Data Requirement
REQUIREMENT
✏️ Open Source MapGPT for current data and Closed Source MapGPT for Future Climate Data and its solutions.
✏️ MapGPT & Solutions - 120k tokens, 2100 Years for Predictive Modeling. Spatial Window for a country like the US.
✏️ GPT-4/BERT, Google Earth Engine, Google Maps API, ESRI ArcGIS API, Global Weather stations, Remote sensing data, Census, Research papers
✏️ Fine tuning GPT-4 on geospatial data sources and context windows with related queries.
✏️ There will be a real-time geoprocessing of certain datasets which will be automated from the text response on the Chatbot. This data processing could be done earlier as well instead of real-time.
✏️ Size will include the full GPT-4 + trained data sources for geospatial data from APIs or databases.
SPECIFICATION
Open vs Closed Source
Context Window
Modalities
Fine-Tuning Capability
Latency
Size
RATIONALE
💡MapGPT is an LLM such as GPT-4 that is linked with an API or a Cloud Geodatabase with Spatial Data Files. (Ex: ArcGIS Living Atlas or IPCC Global Climate Data or US census). The Solutions are linked repository of all the solutions which have a carbon capture logic.
💡MapGPT will be using a context window of 120k tokens to understand user queries using NLP and link them to geodatabases, APIs and spatial data as outputs. The predictive climate modeling is done for 2100 years.
💡GPT-4 will handle Natural Language tasks and link to data sources for geospatial climate data with geospatial APIs & Earth Engine. This can be done using embeddings for geospatial data and combining them with text embeddings.
💡This allows to train they LLM with domain specific knowledge and terminologies. It also will have a feedback and learning loop.
💡LLM will search the geodatabase for the text-context specific data and will display it as a visualization.
💡The size will increase based on how many cities we include in the MVP. The aim is to make it global eventually.
Scrum Plan for Engineering Sprints
Product Manager: Responsible for product and customer issue and prioritizing product backlog items
Scrum Master: Facilitates scrum process and removes impediments
UX Designer: Responsible for UX Research and UX and UI Design of the Product
Marketing Manager: Responsible for GTM strategy and overall marketing
Development Team: Develops and tests features and functionality
LLM Prompt Requirements
You are a GIS analyst, you can find information of climate geospatial data and you will analyze information from the user and generate geospatial data on the maps interface which is downloadable and you will be able to generate graphs that analyze the map data.
Human Evaluation & Risk-Mitigation
❓ Evaluation Question 1: Is the data generated correct? (Y/N)
❓ Evaluation Question 2:Is the data generated accurate? (1-5)
❓ Evaluation Question 3: Is the data generated on the correct site/ context? (Y/N)
❓ Evaluation Question 4: Is the data generated with the right source? (Y/N)
❓ Evaluation Question 5: How specific does the user have to be for the data? (1-5)
Risks
-
The data generated is not correct.
-
The data generated is not granular.
-
The data is off-color.
-
The data is not trustable.
Mitigations
-
List relevant sources and have a feedback learning loop.
-
Tell the user it’s the most granular data that’s available
-
Allow users to change colours and legend items
-
Give relevant data sources and disclaimers
Go to Market Strategy
Message: "Empower decisions with Climate Change Data"
Launch Owner
Target Market
Monetization Strategy
Marketing Strategy
Campaign Effectiveness
Sales Strategy
Support & Maintenance
Dhruv Vairagi
Architecture, Real Estate, City Governments, Urban Planners, Construction Companies, Insurance Providers
Tiered Subscription Model - Basic, Intermediate and Advance
Enterprise Discounts Strategy
API integration & monetization
Content Marketing (4 blog posts/ week and 6 webinars/ year, 3k social media followers, 10 influencers)
Paid Media ($5000 on monthly ad spend in the first year )
Public Relations (2 media outreach per month and 1 thought leadership piece per quarter)
CLV (Customer Lifetime Value)
Website Traffic and social media metrics
Lead generation and conversion rate;
ROI and growth rate
20 demos and 10 trials per month in the first year
10 reseller agreements in channel partnerships in the first year
New marketing campaign 3 months after launch
30% increase in growth rate and 500+ users in the first year
Reduction in Customer Support 3 months after launch
Learnings
This project served as a big learning curve for me. What I learnt was the possibility that AI can achieve for products and how climate change issues can be addressed with ethical AI usage.
Future Scope:
1. Definitely want to make this more feasible with lesser API integrations
2. Have more climate risks integrated
3. Integrate different cities' data
