Africa's Farmers Deserve Intelligence, Not Exploitation.
HarvestIQ delivers real-time AI price intelligence and spoilage prediction to Nigerian smallholder farmers — via USSD, SMS, and voice. No smartphone. No internet. No middleman.
of sub-Saharan African food is lost post-harvest due to lack of buyer coordination and spoilage timing.
📉
$9 Billion
annual post-harvest loss in Nigeria alone, destroying rural wealth before it reaches the market.
🛡️
30–60%
below true market value is what farmers earn, actively manipulated by opaque middleman pricing.
Nigeria alone loses ₦6.5 trillion worth of food annually. This is not a farming failure — it is an information failure. Middlemen arrive at farm gates armed with fabricated price data. Farmers, racing against their produce's biological clock, sell for fractions of true value.
"The middleman knows the price. The farmer doesn't. HarvestIQ closes that gap."
Access Layers
Four Channels. Zero Barriers.
Built for the reality of rural African infrastructure.
📞
USSD (*384*7#)
Any basic phone. Zero data. Under 3 seconds. A farmer in Nasarawa gets current prices for his yam at 5 markets — no app, no internet, no literacy required.
💬
SMS Alerts
Every morning at 7am, before middlemen arrive, HarvestIQ pushes a personalised price bulletin with current price, 3-day trend, and a sell/wait recommendation.
🎙️
Voice IVR
Automated voice in Hausa, Yoruba, Igbo, and Pidgin reads prices aloud. Press 1 for today's price. Press 3 to connect to a verified buyer. Zero literacy required.
👤
Field Agent App
One trusted person in every farming community manages 50 farmers, relays AI recommendations in local dialect, and earns 5% commission on every transaction.
No smartphone required. No credit card. Just a phone number.
Core Technology
The Brain Behind the Farm.
HarvestIQ processes millions of data points across Nigerian markets, weather systems, and biological decay models to generate a single, actionable sentence for a farmer.
⚙️
Spoilage Prediction Model
Trained on 200+ produce samples across Nigerian temperature and humidity ranges. Calculates degradation timeline for 8 crop types and outputs a countdown timer to Grade C/unsellable.
Real-Time Price Intelligence
Continuously ingests live price feeds from Nigeria's 4 largest terminal markets. Our Amazon Forecast model processes 72 hours of rolling price history with weather correlations and supply-demand signals to generate 7-day forecasts with 85%+ directional accuracy.
Spoilage Window Prediction
SageMaker-hosted biological decay model calculates the exact degradation timeline for 8 crop types. Output: a countdown timer with quality grade decay stages — Grade A → B → C → unsellable — plus cold storage recommendations.
Optimal Sell Decision Engine
HarvestIQ's core IP merges the price forecast curve with the spoilage window countdown to compute the optimal sell date and market — maximising farmer earnings before quality drops below Grade A.
Direct Buyer Connection
A verified, rated directory of produce buyers indexed by crop type, quantity, market location, and payment speed. HarvestIQ connects farmers directly with escrow-style payment protection.
Logistics Optimisation
Route recommendation engine using Amazon Location Service — real-time road conditions, transit time vs. spoilage window comparison. Group logistics coordination for cooperatives sharing transport.
Computer Vision Quality Grading
Field agents photograph produce and receive instant AI quality grading via Amazon Rekognition. Buyers get objective quality data before purchase, enabling premium pricing for Grade A produce.
AWS Partner
How We Use Amazon Web Services
Every layer of HarvestIQ runs on AWS. Here is exactly how.
Amazon SageMaker
We train and host two core AI models: our spoilage window prediction model (regression-based, trained on biological decay data across 8 crops) and our price forecasting model (time-series, 6 months of Nigerian market data). SageMaker endpoints respond in under 800ms — fast enough for USSD where every second costs the farmer network fees.
Amazon Forecast
Powers our 7-day market price prediction. Ingests daily price data from 4 terminal markets, applies AutoML to select the optimal forecasting algorithm, and outputs directional predictions (rising/stable/falling) that feed directly into our sell decision engine.
Polly + Transcribe + Connect
Polly converts AI recommendations into speech in 4 languages: Hausa, Yoruba, Igbo, and Nigerian Pidgin — the only agricultural voice AI in West Africa operating in local languages. Amazon Connect manages inbound/outbound call infrastructure.
Kinesis + Lambda + API Gateway
Real-time price data ingestion runs on Kinesis Data Streams, processed by Lambda every 30 minutes. API Gateway handles USSD webhooks from telco partners (MTN, Airtel), SMS callbacks from Africa's Talking, and WhatsApp Business API events.
RDS + DynamoDB + S3
RDS PostgreSQL hosts farmer profiles, subscription records, transaction history, and buyer directory. DynamoDB stores real-time USSD session state (sub-10ms latency). S3 stores voice recordings, produce photographs, training datasets, and field agent app assets.
Rekognition + Comprehend
Rekognition powers produce quality grading — field agents photograph tomatoes, peppers, or yams and receive instant Grade A/B/C classification. Comprehend processes incoming text queries in English and local languages, routing to the correct AI pipeline.
CloudFront + Route 53 + Security
CloudFront delivers the Field Agent mobile app and web dashboard with sub-50ms latency across Nigeria. GuardDuty provides continuous threat detection. KMS encrypts all farmer data — NDPR compliant.
EC2 Auto Scaling + Amplify
EC2 Auto Scaling handles USSD processing, scaling from 2 to 50 instances during peak harvest seasons. Amplify hosts our Field Agent React Native app with offline sync capability — critical for areas with intermittent connectivity.
NVIDIA Technology
How We Use NVIDIA Technology
GPU acceleration makes our AI models fast enough for real-time USSD interactions.
NVIDIA CUDA & cuDNN
Our spoilage prediction and price forecasting models require GPU-accelerated training to iterate on new crop data efficiently. CUDA enables parallel computation — what takes 8 hours of CPU training completes in under 40 minutes on a GPU. As we expand to 8 crops and 10,000+ farmers, GPU training becomes non-negotiable for model refresh cadence.
NVIDIA TensorRT
For USSD inference speed is critical. Farmers dial in with a 180-second timeout. Our SageMaker endpoints, optimised with TensorRT, deliver spoilage predictions in under 800ms. TensorRT reduces model size by up to 4x through quantisation while maintaining accuracy, enabling cost-efficient production deployment.
NVIDIA Jetson (Edge AI)
Phase 2 includes solar-powered edge devices deployed at cold storage facilities and market checkpoints. Jetson Nano modules will run produce quality grading inference locally, without cloud connectivity, enabling real-time Grade A/B/C classification for farmers too far from a mobile network.
Computer Vision Pipeline (CUDA)
As our training dataset grows beyond 100,000 produce images, we plan to train a custom vision model using CUDA-accelerated PyTorch on NVIDIA A10G GPUs via SageMaker — outperforming the general-purpose Rekognition model for Nigerian produce varieties and local quality standards.
Market Opportunity
A $400 Billion Problem Across a Continent
$50B+
Nigerian agriculture GDP contribution
38M
Smallholder farmers in Nigeria alone
85%
Of food consumed is produced by smallholders
The infrastructure to fix this cannot wait for nationwide 5G or smartphone adoption. It must work today, on the networks that exist, on the devices farmers already own.
Founding Team
Combining supply chain expertise with deep technical execution.
Ifunanya Ukwueze
Co-Founder & CEO
8 years in Nigerian FMCG supply chain. Former analyst at Dangote Group. MBA, Lagos Business School. Identified the post-harvest intelligence gap after watching farmers sell tomatoes at ₦15,000/tonne on the same day Mile 12 buyers were paying ₦45,000.
Aisha Bello
Co-Founder & CTO
AWS Certified Solutions Architect. Former senior engineer at Andela. BSc Computer Science, University of Lagos. Built HarvestIQ's USSD backend and SageMaker inference pipeline. Speaker at PyCon Africa 2024.
Every farmer deserves to know what his harvest is worth before the middleman tells him.
