{"success":true,"data":{"data":[{"id":"model-001","name":"Soil nutrient prediction (geospatial feature engineering)","shortDescription":"Feature-rich geospatial pipeline (1,000+ engineered features) + soil-depth matching to predict multiple nutrients from satellite/climate/soil signals.","fullDescription":"Soil-nutrient prediction solution (Rhea challenge) built around heavy feature engineering and domain knowledge. Key elements: (1) 1,000+ features derived from satellite, climate and soil datasets, (2) soil-depth matching to align observations and improve signal consistency, and (3) multi-target regression to estimate several nutrient values at unsampled points. Intended for scalable nutrient mapping to guide fertiliser recommendations and soil-health interventions.","useCase":"Estimate multiple soil nutrients from geospatial + soil data at locations without lab tests.","technicalRequirements":"Uses open-source tools; may incorporate earth observation + climate datasets; must include preprocessing + download scripts for any external data.","sector":"climate-resilience","country":"KE","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/Rhea/Solutions_SHARED","hackathonId":"hack-kigali-2025"},{"id":"model-002","name":"Soil nutrient prediction (domain-aware ensemble)","shortDescription":"Domain-aware ensemble that blends Random Forest + LightGBM, with adversarial feature selection; separate inference path for historical vs 2025 data.","fullDescription":"Soil-nutrient prediction solution (Rhea challenge) designed for robust generalisation across locations. Key elements: (1) domain-aware ensemble combining Random Forest and LightGBM, (2) adversarial feature-selection strategy to reduce spurious predictors, and (3) a split inference pipeline handling historical vs 2025 data differently to mitigate distribution shift. Outputs: multi-target regression for several soil nutrients, enabling nutrient mapping where lab tests are unavailable.","useCase":"Estimate multiple soil nutrients from geospatial + soil data at locations without lab tests.","technicalRequirements":"Uses open-source tools; may incorporate earth observation data (e.g., Sentinel-2, Landsat) and climate datasets; must include preprocessing + download scripts for external data if used.","sector":"climate-resilience","country":"KE","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/Rhea/Solutions_SHARED","hackathonId":"hack-kigali-2025"},{"id":"model-003","name":"Soil nutrient prediction (climate-zone & village models)","shortDescription":"Climate-zone + village-level modelling: Köppen–Geiger clustering, hierarchical statistics, and an ensemble to improve nutrient estimates without lab tests.","fullDescription":"Soil-nutrient prediction solution (Rhea challenge) emphasizing spatial generalisation. Approach: (1) clusters samples by climate similarity using Köppen–Geiger zones, (2) builds village-level models with hierarchical statistics to capture local patterns, and (3) combines diverse learners in an ensemble for stable multi-nutrient regression. Designed to support tailored soil recommendations where laboratory analysis is limited.","useCase":"Estimate multiple soil nutrients from geospatial + soil data at locations without lab tests.","technicalRequirements":"Uses open-source tools; may incorporate earth observation + climate datasets; must include preprocessing + download scripts for any external data.","sector":"climate-resilience","country":"KE","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/Rhea/Solutions_SHARED","hackathonId":"hack-kigali-2025"},{"id":"model-004","name":"Customer retention (LTV/retention modelling pipeline)","shortDescription":"Upload coming soon","fullDescription":"Upload coming soon","useCase":"Upload coming soon","technicalRequirements":"Python ML workflow for agri-ecommerce customer intelligence (pandas/numpy, scikit-learn). Supports segmentation (clustering) + retention/LTV modelling; includes preprocessing + feature engineering; reproducible training/inference code; dependency file and run steps.","sector":"supply-chain","country":"GH","githubUrl":null,"hackathonId":"hack-accra-2025"},{"id":"model-005","name":"Customer retention (RFM features + supervised churn/propensity)","shortDescription":"Upload coming soon","fullDescription":"Upload coming soon","useCase":"Upload coming soon","technicalRequirements":"Customer segmentation + retention modelling stack (Python: pandas/numpy, scikit-learn + optional XGBoost/LightGBM). Includes preprocessing of customer/activity/order data, feature engineering (RFM/cohorts), clustering or supervised propensity/churn models, and reproducible training + inference scripts; dependency file + run instructions.","sector":"supply-chain","country":"GH","githubUrl":null,"hackathonId":"hack-accra-2025"},{"id":"model-006","name":"Customer retention (customer segmentation via clustering + cohorts)","shortDescription":"Upload coming soon","fullDescription":"Upload coming soon","useCase":"Upload coming soon","technicalRequirements":"Python analytics/ML pipeline for segmentation & retention (pandas/numpy, scikit-learn + optional boosting). Implements data cleaning, feature engineering, segmentation (clustering) and/or retention prediction; reproducible notebooks/scripts; dependency file + README.","sector":"supply-chain","country":"GH","githubUrl":null,"hackathonId":"hack-accra-2025"},{"id":"model-007","name":"Produce recommendations (baseline/shared solution)","shortDescription":"Baseline/shared recommender system solution for the Farm to Feed online challenge (Zindi).","fullDescription":"Baseline/shared solution produced for the AI4SU online hackathon “Produce Recommendation Systems for Food Waste Reduction” (Farm to Feed, hosted on Zindi). The goal is to build a recommender system using anonymised transaction data to (1) predict purchase likelihood in the next 7 and 14 days and (2) estimate expected quantities per customer and SKU. This supports more relevant recommendations, better inventory and marketing decisions, and ultimately helps reduce food waste while improving access to fresh produce markets.","useCase":"Recommend fresh produce to customers by predicting which items they are likely to purchase in the next 7 and 14 days, and estimating expected quantities—helping reduce food waste and improve e-commerce operations for Farm to Feed.","technicalRequirements":"Recommender system / next-basket prediction in Python (pandas/numpy, scikit-learn + optional implicit/LightFM or gradient boosting). Must output purchase likelihood for 7/14 days and quantities per customer/SKU; includes transaction preprocessing, feature engineering, and reproducible inference scripts. Dependency file + run instructions.","sector":"supply-chain","country":"KE","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/FarmToFeed/Solutions_SHARED","hackathonId":"hack-cairo-2025"},{"id":"model-008","name":"Climate & market resilience (solution 1)","shortDescription":"Hackathon model: climate & market resilience (FarmSawa).","fullDescription":"Model produced for the FarmSawa hackathon challenge “Climate & Market Resilience” (Nairobi, Kenya). Focus: resilience to climate and market volatility for smallholder farmers.","useCase":"Anticipate climate shocks and market volatility to protect smallholder farmers’ incomes.","technicalRequirements":"Python ML stack for tabular modelling (pandas/numpy, scikit-learn + optional LightGBM/XGBoost). Requires data preprocessing, feature engineering, model training + inference scripts/notebook; dependency file (requirements.txt/conda) and reproducible run instructions. (Repo: Ai4SU)","sector":"climate-resilience","country":"KE","githubUrl":"https://github.com/MichaelgGB/Ai4SU","hackathonId":"hack-nairobi-2025"},{"id":"model-009","name":"Credit scoring (hybrid index: satellite + mobile data)","shortDescription":"Hackathon model: inclusive credit scoring (FarmSawa).","fullDescription":"Model produced for the FarmSawa hackathon challenge “Inclusive Credit Scoring for Smallholder Farmers” (Nairobi, Kenya). Focus: inclusive credit scoring and financial inclusion.","useCase":"Unlock access to agricultural finance for farmers without formal credit histories.","technicalRequirements":"Python credit-scoring ML pipeline (pandas/numpy, scikit-learn + optional LightGBM/XGBoost). Data cleaning + feature engineering, model training, and inference for creditworthiness scoring; handling missing/categorical values; dependency file + reproducible run instructions.","sector":"agri-finance","country":"KE","githubUrl":"https://github.com/GreenVast","hackathonId":"hack-lagos-2025"},{"id":"model-010","name":"Climate & market resilience (solution 2)","shortDescription":"Hackathon model: climate & market resilience (FarmSawa).","fullDescription":"Model produced for the FarmSawa hackathon challenge “Climate & Market Resilience” (Nairobi, Kenya).","useCase":"Anticipate climate shocks and market volatility to protect smallholder farmers’ incomes.","technicalRequirements":"Python ML stack (pandas/numpy, scikit-learn + optional LightGBM/XGBoost). Includes preprocessing + feature engineering, model training and inference code; clear instructions to reproduce results and run predictions; dependency file.","sector":"climate-resilience","country":"KE","githubUrl":"https://github.com/geosatt/GeoSatt-Hub-AI1","hackathonId":"hack-nairobi-2025"},{"id":"model-011","name":"Credit scoring (XAI risk dashboard & fairness)","shortDescription":"Hackathon model: inclusive credit scoring (FarmSawa).","fullDescription":"Model produced for the FarmSawa hackathon challenge “Inclusive Credit Scoring for Smallholder Farmers” (Nairobi, Kenya).","useCase":"Unlock access to agricultural finance for farmers without formal credit histories.","technicalRequirements":"Python tabular ML stack for inclusive credit scoring (pandas/numpy, scikit-learn + optional boosting). End-to-end preprocessing + training + inference code; produces credit risk/score outputs; dependency file and clear run steps.","sector":"agri-finance","country":"KE","githubUrl":"https://github.com/Team-Bombadier","hackathonId":"hack-lagos-2025"},{"id":"model-012","name":"Climate & market resilience (solution 3)","shortDescription":"Hackathon model: climate & market resilience (FarmSawa).","fullDescription":"Model produced for the FarmSawa hackathon challenge “Climate & Market Resilience” (Nairobi, Kenya).","useCase":"Anticipate climate shocks and market volatility to protect smallholder farmers’ incomes.","technicalRequirements":"Python data/ML workflow (pandas/numpy, scikit-learn + optional boosting). Must include reproducible preprocessing and training pipeline, plus inference script to generate predictions; dependency file and run steps in README.","sector":"climate-resilience","country":"KE","githubUrl":"https://github.com/Gill-tech/M-nda-AI","hackathonId":"hack-nairobi-2025"},{"id":"model-013","name":"Credit scoring (behavior-based predictive model)","shortDescription":"Hackathon model: inclusive credit scoring (FarmSawa).","fullDescription":"Model produced for the FarmSawa hackathon challenge “Inclusive Credit Scoring for Smallholder Farmers” (Nairobi, Kenya).","useCase":"Unlock access to agricultural finance for farmers without formal credit histories.","technicalRequirements":"Python ML credit scoring workflow (pandas/numpy, scikit-learn + optional LightGBM/XGBoost). Includes preprocessing, feature engineering, training + inference scripts; robust handling of missing values; dependency file + README instructions.","sector":"agri-finance","country":"KE","githubUrl":"https://github.com/FARMIQ-T","hackathonId":"hack-lagos-2025"},{"id":"model-014","name":"Produce recommendations (next-basket prediction, female team)","shortDescription":"Winning/shared solution: produce recommendations (Farm to Feed).","fullDescription":"Winning/shared solution from the Farm to Feed Zindi challenge “Produce Recommendation Systems for Food Waste Reduction” (Kenya).","useCase":"Predict future shopping baskets to improve fresh produce recommendations (7/14 days + quantities).","technicalRequirements":"Python ML stack for basket prediction (pandas/numpy, scikit-learn + optional LightGBM/XGBoost). Produces 7/14-day purchase probabilities and item quantity forecasts; engineered features from historical transactions; handles categorical IDs and sparsity; reproducible training/inference code + dependency file + readme.","sector":"supply-chain","country":"KE","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/FarmToFeed/Solutions_SHARED/2nd_Top_Female_Team_Solution","hackathonId":"hack-cairo-2025"},{"id":"model-015","name":"Practice adoption prediction (structured preprocessing for generalisation)","shortDescription":"Practice-adoption model tuned for Kenyan-context generalisation: structured preprocessing + supervised learning for 7/90/120-day adoption probabilities.","fullDescription":"Top Kenyan team solution for the DigiCow Zindi challenge “Predicting Practice Adoption in Smallholder Farming”. Uses a structured preprocessing + modelling approach to estimate adoption probabilities at 7, 90 and 120 days from farmer training data, supporting targeted extension follow-ups.","useCase":"Predict which farmers will adopt practices after training (within 7/90/120 days).","technicalRequirements":"Python ML stack (pandas/numpy, scikit-learn + boosting library). Preprocessing for training-time variables only; produces probability outputs for 7/90/120 days; reproducible run instructions + dependency file; robust handling of missing/categorical fields.","sector":"livestock","country":"KE","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/DigiCow/Solutions_SHARED/3rd_Top_Kenyan_Team_Solution","hackathonId":"hack-dakar-2025"},{"id":"model-016","name":"Practice adoption prediction (end-to-end feature engineering + classification)","shortDescription":"End-to-end practice-adoption predictor (7/90/120 days): feature engineering on training records + classification models with probability outputs.","fullDescription":"Solution for the DigiCow Zindi challenge “Predicting Practice Adoption in Smallholder Farming”. Implements an end-to-end pipeline (data cleaning, feature engineering from training-time variables, model training and inference) to estimate adoption probabilities at 7, 90 and 120 days. Designed to help prioritise follow-ups and tailor extension support.","useCase":"Predict which farmers will adopt practices after training (within 7/90/120 days).","technicalRequirements":"Python ML stack (pandas/numpy, scikit-learn and/or gradient boosting like LightGBM/XGBoost). Reproducible training + inference scripts/notebooks; outputs calibrated probabilities for 7/90/120-day adoption; handles missing values and categorical encoding; requirements.txt/environment file.","sector":"livestock","country":"KE","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/DigiCow/Solutions_SHARED/Top_Overall_Solution","hackathonId":"hack-dakar-2025"},{"id":"model-017","name":"Practice adoption prediction (robust probability estimation)","shortDescription":"Practice-adoption prediction pipeline (7/90/120 days) optimized for stable probability estimates and leaderboard performance (female team award).","fullDescription":"Awarded solution for the DigiCow Zindi challenge “Predicting Practice Adoption in Smallholder Farming”. Provides a reproducible modelling workflow to predict adoption at 7/90/120 days using only training-time information, with attention to robust probability estimation for targeting interventions and monitoring adoption.","useCase":"Predict which farmers will adopt practices after training (within 7/90/120 days).","technicalRequirements":"Python ML stack (pandas/numpy, scikit-learn and/or LightGBM/XGBoost/CatBoost). End-to-end pipeline with preprocessing, feature engineering, and probability prediction for 7/90/120 days; reproducible notebook/script + dependency file; clear train/validation split and inference entrypoint.","sector":"livestock","country":"KE","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/DigiCow/Solutions_SHARED/2nd_Top_Female_Solution","hackathonId":"hack-dakar-2025"},{"id":"model-018","name":"Produce recommendations (next-basket prediction, end-to-end)","shortDescription":"Winning/shared solution: produce recommendations (Farm to Feed).","fullDescription":"Winning/shared solution from the Farm to Feed Zindi challenge “Produce Recommendation Systems for Food Waste Reduction” (Kenya).","useCase":"Predict future shopping baskets to improve fresh produce recommendations (7/14 days + quantities).","technicalRequirements":"Python recommender/forecasting stack (pandas/numpy, scikit-learn + optional LightGBM/XGBoost). Must support (1) binary classification for purchase likelihood at 7 and 14 days and (2) regression for quantities per customer/SKU. Includes feature engineering on transaction history (recency/frequency, basket signals), handling sparse user×item matrices, reproducible training + inference scripts, dependency file, and clear run instructions.","sector":"supply-chain","country":"KE","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/FarmToFeed/Solutions_SHARED/1st_Top_Solution","hackathonId":"hack-cairo-2025"},{"id":"model-019","name":"Maize price forecasting (lag/rolling features pipeline)","shortDescription":"County-level maize price predictor (Kenyan award): lag/rolling features + ML regression for weekly price forecasts.","fullDescription":"3rd top Kenyan solution for the agriBORA Zindi challenge “Maize Price Intelligence for Smarter Selling”. Implements a forecasting pipeline for weekly maize prices across Kenyan counties using engineered time-series features and regression modelling to support price intelligence.","useCase":"Forecast average weekly maize prices across Kenyan counties to support sell-timing and market intelligence.","technicalRequirements":"Python forecasting pipeline (pandas/numpy, scikit-learn + boosting). Builds lag/rolling features; generates county-level weekly predictions; reproducible training/inference code with requirements.txt/environment; clear run steps.","sector":"supply-chain","country":"KE","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/agriBORA/Solutions_SHARED/3rd_Top_Kenyan_Solution_Owino/klerk002","hackathonId":"hack-johannesburg-2025"},{"id":"model-020","name":"Crop type mapping (pixel classification, solution 1)","shortDescription":"Winning/shared solution: pixel-level crop classification (Tolbi).","fullDescription":"Winning/shared solution from the Tolbi Zindi challenge “Pixel-Level Crop Classification for Precision Agriculture” (Côte d’Ivoire).","useCase":"Classify crop type at pixel level (cocoa/rubber/oil palm) using Sentinel-2 imagery for scalable crop mapping.","technicalRequirements":"Multi-temporal satellite image modelling for pixel-level crop mapping. Python + geospatial preprocessing (rasterio/xarray) and ML/DL (scikit-learn and/or PyTorch/TensorFlow). Must manage cloud/missing values, normalise bands, and output pixel labels; reproducible training/inference code with dependency file.","sector":"crop-science","country":"CI","githubUrl":"https://github.com/ZindiAfrica/Digital-Africa/tree/main/Tolbi/Solutions_SHARED/Winning%20Solutions/%5C%20Les%20Elephants","hackathonId":"hack-casablanca-2025"}],"total":24,"page":1,"limit":20,"totalPages":2}}