- Project summary
- Technology implementation
- Judging criteria
- Featured software products
- Github repository
Climate change is intensifying risks to infrastructure, agriculture, and vulnerable communities. Engineers, researchers, insurers and policymakers increasingly require integrated, data-driven systems that can assess resilience, model correlated climate hazards and trigger timely, accountable responses when risk thresholds are exceeded.
Current approaches rely on siloed tools: AutoCAD for design, Excel for statistical analysis, standalone climate databases for hazard records and manual decision-making for interventions. These workflows are slow, error-prone, and poorly suited to dynamic climate risk.
This project proposes a MeTTa-based Climate Resilience Platform that unifies heterogeneous data into a digital twin–driven, rule-based decision system, augmented—but not replaced—by large language models (LLMs). The platform enables explainable engineering calculations, climate risk reasoning, and automated yet governable responses such as insurance payouts, disaster relief activation, and adaptive infrastructure design.
MeTTa is used as a bridging language to interact with OpenAI's models through a framework called MeTTa-Motto OpenAI models are used as constrained interpretive and synthesis components within a MeTTa-governed climate resilience platform. They provide explanation, summarization, and code assistance under explicit logical and functional boundaries, while all safety-critical reasoning, thresholds, and automated actions remain governed by formal MeTTa rules and human oversight. OpenAI’s models play a supporting, bounded, and explainable role within the system. They are not used as primary decision-makers, but as intelligent components invoked through MeTTa-Motto under strict logical and functional constraints. MeTTa-Motto allows developers to use OpenAI's GPT-4 or other LLMs as "components" within a MeTTa script It enables the calling of functions and the generation of JSON responses from OpenAI models that adhere to specific signatures defined in MeTTa.
The MeTTa programming language offers a novel way to unify heterogeneous data from AutoCAD dimensions, Excel datasets and environmental hazard records into logical, mathematical and semantic models that can power smart contracts and adaptive climate responses.
Assessment of dimensions (AutoCAD)
Problem: Engineers often manually calculate flow rates, pipe volumes, and plant capacities from drawings, which is slow and error-prone.
MeTTa Solution: A programmatic way to take AutoCAD parameters (length, diameter, velocity) and automatically compute water flow, greenhouse insulation, and plant capacity for climate-resilient infrastructure.
Mapping correlations of disasters
Problem: Climate events (floods, droughts, earthquakes) are logged separately, without a unified system for identifying patterns and community risk.
MeTTa solution: Assigns unique identifiers to each region and event, computes risk indices, and reveals disaster correlations to guide interventions (e.g., flood-prone regions get reinforced irrigation networks).
Triggering Smart Contracts (Excel + MeTTa)
Problem: Statistical thresholds from Excel (e.g., rainfall deviations > 2σ from mean) are not directly tied to financial or insurance responses.
MeTTa Solution: Embeds these metrics into if-then rules, enabling automatic triggering of insurance payouts, disaster relief funds, or supply-chain adjustments via blockchain smart contracts.
The solution could help by gathering real time data so as to aid reseachers such as geologists, agriculturalists and water engineers to make appropriate decisions with regard to designing agricultural equipment, formaulating farming practices, accurately analysing weather patterns, creation of code that relates to medical prescription and providing explanation to code. The technology helps farmers in mitigating effects of climate change that affects crops, the goal is to conduct an assessment of where farmers do farming then in the event of extreme climate conditions such as drought or flooding data is collected which provides action on what remedial steps to be taken. The assessment is done through farmers taking pictures and other relevant information is collected which is analysed using software such as AutoCAD and other analytical software from IBM. Codes are also generated where they are needed by use of MeTTa programing language, all the outputs are finally sent to Open AI to conduct analysis. Reviews are also taken from farmers based on how Afterwards, all the information is collected and stored as a refernce for future analysis.
As a short demo, this link can be used by African communities to upload photos or videos. Additionally, technical experts can use the data obtained from African communities to derive insights and eventually send reports to disaster management firms and insurance companies https://climate-witness-repo-nneg.bolt.host/
The idea is creation of a dataset comprised of images mostly from local farmers and other detailed information such as Excel sheets and documents from agricultural experts. This data will be useful to train Open AI models through use of programming language such as MeTTa or even python as well as use of software such as AutoCAD and even other analytical software from IBM and many others. The data obtained especially images are analysed in detail by assessment of feature such as dimension, color, geographical location and other factors with the goal of deriving as much insights as possible.
The central innovation is the use of the MeTTa programming language as a formal semantic and logical layer that:
Integrates geometric, statistical, and environmental data
Encodes deterministic engineering laws and probabilistic risk models
Governs when and how LLMs, through MeTTa-Motto, are invoked
Produces auditable, explainable outputs suitable for high-stakes decisions
Through MeTTa-Motto, OpenAI models are used as constrained components for interpretation, explanation, and synthesis—never as opaque decision-makers.
Step 1: AutoCAD integration
Export design attributes especially for farmers who are the most impacted by climatic disasters. The aspects being examined include pipe diameters, greenhouse cover thickness, plant footprint into structured data.
MeTTa computes real-time engineering metrics (flow rate, material insulation, plant capacity).
Step 2: Excel integration
Climate and economic datasets stored in Excel sheets (rainfall, yield, disaster costs).
MeTTa queries Excel data for statistical analysis (mean rainfall, variance in crop yield).
Step 3: Disaster correlation engine
Disaster events (flood, drought, earthquake) are logged with region, date, severity.
MeTTa calculates risk indices per region and generates community IDs.
Step 4: Smart contract triggers
When thresholds are breached such as MeTTa outputs conditions to a blockchain system.
Smart contracts auto-execute this includes release of microinsurance payouts, allocation of emergency water supply, or subsidy distribution
Relevance of MeTTa
Technical implementation
Usability and user experience
Impact and scalability 
AutoCAD → Provides engineering drawings and dimensional data of assets such as pipes, greenhouses, and irrigation systems. These dimensions illustrate parameters such as capacity, flow rate and even material requirements
Microsoft Excel → Performs baseline statistical calculations such as mean, variance, standard deviation from climate and economic datasets, enabling detection of anomalies and performance trends.
MeTTa → Integrates AutoCAD-derived measurements and Excel-calculated metrics into a knowledge graph, where correlations can be mapped such as droughts versus. crop failures. It further enables automated reasoning and triggering of smart contracts once disaster thresholds are detected. For the Metta file of ProjectAgriAccess the end result is a ratio in the form of a decimal that describes the extent of damage.
(pipe "P1" length 100 diameter 0.5 velocity 2)
(calc-pipe-flow (pipe $id length $l diameter $d velocity $v) (= $area (* 3.1416 (* (/ $d 2) (/ $d 2)))) # Area = πr² (= $flow (* $area $v)) # Flow = Area * velocity (return $flow) )
(greenhouse "GH1" thickness 0.01 material "polyethylene")
(calc-insulation (greenhouse $id thickness $t material $m) (= $insulation (* 200 $t)) # Example formula: factor = 200 * thickness (return $insulation) )
(plant "Plant1" length 0.5 width 0.3 thickness 0.1)
(calc-plant-volume (plant $id length $l width $w thickness $t) (= $volume (* $l $w $t)) # Volume = L * W * T (return $volume) )
(disaster "Flood" region "R1" severity 8 time 2020) (disaster "Drought" region "R1" severity 6 time 2021) (disaster "Earthquake" region "R2" severity 9 time 2022)
(assign-id (disaster $type region $r severity $s time $t) (= $id (+ (* $s 1000) $t)) # ID formula: severity*1000 + year (return $id) )
(map-correlation (disaster $type1 region $r severity $s1 time $t1) (disaster $type2 region $r severity $s2 time $t2) (= $diff (- $t2 $t1)) # Time gap between disasters (= $severity-gap (- $s2 $s1)) # Change in severity (return (list $diff $severity-gap)) )
(excel-data "rainfall" values (100 120 90 80 110))
(calc-mean (excel-data "rainfall" values $vals) (= $sum (reduce + $vals)) # Sum of all values (= $n (length $vals)) # Count of values (= $mean (/ $sum $n)) # Mean = sum/n (return $mean) )
(calc-stddev (excel-data "rainfall" values $vals) (= $mean (calc-mean (excel-data "rainfall" values $vals))) (= $sqdiffs (map (lambda $x (- $x $mean)^2) $vals)) # Squared diffs (= $variance (/ (reduce + $sqdiffs) (length $vals))) (= $stddev (sqrt $variance)) (return $stddev) )
(trigger-smart-contract (excel-data "rainfall" values $vals) (= $stddev (calc-stddev (excel-data "rainfall" values $vals))) (if (> $stddev 15) # Threshold for anomaly (return "Smart Contract Triggered: Insurance payout released") (return "No Trigger: Conditions normal") ) )
In the Github section links for the python code of the use of watsonx.ai has been placed. In the docs section powerpoint of the presentation used in the video has been placed. An excel sheets highlighting contents of the Environmental Intelligence Suite has also been placed.
Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.
We use SemVer for versioning. For the versions available, see the tags on this repository.
<a
- John Maina - Initial work -
This project is licensed under the Apache 2 License - see the LICENSE file for details.