Quorvex Labs · Predictive maintenance for land-reform farm infrastructure

Hear the failure before it breaks the season.

FarmPulse listens to irrigation lines, fences and storage through small vibration, flow and structural sensors. It warns when something is wearing out, drafts a maintenance plan, and helps prepare a funding request. A farm support officer approves every repair.

Charts on this page are illustrative and simulated, not real farm data.

Stage: FarmPulse is an early-stage prototype. This site demonstrates the concept with simulated data; there are no live deployments yet. What FarmPulse does not do: it does not guarantee a failure will be caught, replace a physical inspection, order repairs or spend money by itself. The farmer keeps ownership of their data, shared only with consent.

Why infrastructure fails quietly

A pump bearing wears, a joint leaks, a fence post leans. Without early warning, the first sign is often a dry field, a lost herd or a damaged store, and repair money arrives too late.

Wear beginsNo sensorSmall fault unseenSudden failureCrop or stock lossLate funding requestOfficer acts earlier

Healthy

Vibration, flow and structure readings sit in the normal range.

Wearing

A trend is drifting. Preventive maintenance is suggested.

Likely to fail

A strong warning. A repair work order is drafted for approval.

FarmPulse produces predictions for human review, not guarantees.

How it works

1 Sense

Vibration, flow and structural sensors on pipes, pumps, fences and stores.

2 Predict

Readings and history become a failure-risk score.

3 Inspect

Farmer or officer shares a photo for visual inspection support.

4 Plan

A preventive maintenance plan is drafted.

5 Fund

A funding request is drafted with the evidence.

6 Officer decides

Approve work, inspect first, ask the farmer, or dismiss.

The AI helpers

Failure prediction

Estimates failure risk from sensor trends such as vibration, flow and structural change.

Maintenance planning

Drafts a preventive maintenance plan for the officer to review.

Funding drafts

Prepares a funding request with the supporting evidence. Never submitted automatically.

Visual inspection support

Helps review photos of pipes, fences and stores.

Farm support officer (mandatory)

Reviews every warning, work order and funding request. Nothing is issued without them.

AI models we plan to use

The models below are the ones we are evaluating and building around. Status as checked on 30 September 2026. Availability changes quickly, so confirm in each provider's console before relying on it.

OpenAI: language and reasoning

GPT-6 Astra: flagship for complex reasoning and coding
GPT-6.1 Sol: balance of intelligence and cost
GPT-6 Luna: low-cost, high-volume work
gpt-oss-120b and gpt-oss-20b: open-weight models we could self-host
GPT-5.5 and GPT-5.4: previous generation, kept as fallbacks

Source: OpenAI API model docs.

OpenAI: image

GPT Image 2.5: released 8 September 2026 in ChatGPT; check API availability
GPT Image 2 (gpt-image-2): current API model
GPT Image 1.5 and 1 mini: older, cheaper options

Used for training visuals and inspection guides, not for evidence.

Google Cloud (Gemini Enterprise Agent Platform, formerly Vertex AI)

Gemini 3.8 Flash: newest Flash model, for coding and agent workflows
Gemini 3.7 Flash and 3.5 Flash-Lite: lower-cost tiers
Gemini 3.1 Pro and 3 Deep Think: harder reasoning
Imagen 4 (Generate, Fast, Ultra): image generation
Veo 3.1 (video) and Lyria 3 (music, preview)
Gemma open models, and Model Garden with 200+ partner and open models

Source: Google Cloud model docs.

AWS: Amazon Bedrock

Amazon Nova 2 Lite, Nova 2 Sonic (speech) and Nova Multimodal Embeddings
Titan Text Embeddings V2 for search over farm records
Stability AI image models for image generation
OpenAI GPT-6 family on Bedrock, reported from 8 and 22 September 2026
Anthropic Claude, Meta Llama, Mistral, Cohere, DeepSeek and more: 100+ models

First-generation Nova, Nova Canvas and Nova Reel are being retired, so we will not build on them. Source: AWS Bedrock docs and a third-party catalogue.

AWS: other services

Amazon SageMaker: train our own failure-prediction models on sensor data
Amazon Rekognition: custom image checks for pipes, fences and stores
AWS IoT Core: connect field sensors securely

These are standard AWS services, not new models.

Where each helper could use them

Failure prediction: our own time-series model on SageMaker, with an LLM to explain the result
Maintenance plans: GPT-6.1 Sol, Gemini 3.8 Flash or Nova 2 Lite
Funding drafts: GPT-6 Astra or Gemini 3.1 Pro
Photo inspection: multimodal models from any of the above

Chosen by accuracy, cost, data protection and regional availability.

Status. These are candidates under evaluation. None is in production and none processes real farm data today. Listing a model is not a partnership or endorsement, and Quorvex Labs is not affiliated with these providers. Model names are trademarks of their owners.

Responsible AI. A human decides every action. AI outputs on this site are simulated. The final models and cloud provider have not been chosen yet, and will be chosen on accuracy, cost, data-protection terms and South African data requirements. Before real use we will test sensors on real assets, measure false alarms and missed detections, and obtain farmer consent over data.

3D model

Wear and vibration level: 20 — HEALTHY

Research and evidence

Published commentary and research on why land-reform projects struggle. We cite these for context, not as proof that FarmPulse works.

Daily Maverick: Presidential Advisory Panel on Land Reform and Agriculture report (2019)

Vumelana Advisory Fund on post-settlement support funding (2021)

University of Pretoria: revisiting land reform beneficiaries and project outcomes

Farmer's Weekly: slow land reform and poor post-settlement support

Common thread in these sources: weak post-settlement support is repeatedly named as a factor in project failure. FarmPulse targets one part of that gap, infrastructure upkeep. Company research on asset failure rates will be added once pilot data exists.

Where we could start

Real weather from Open-Meteo, refreshed every 10 minutes. Stage labels are the company's own designations.

About Quorvex Labs

Company details

Quorvex Labs (Pty) Ltd · Private company
Registration no.: 2026/723698/07 · Registered with the CIPC: 9 September 2026
Status: In business · Financial year end: March
Registered office: 34 Squirrel Street, Villa Liza, Boksburg, Gauteng, 1459, South Africa
Directors: 1 (Chantelle Palesa Malahlela) · Product: FarmPulse

Chantelle Palesa Malahlela

Founder and Director · chantelle@quorvexlabs.co.za · 081 324 0094

Raised on a farm in the Eastern Cape. Building FarmPulse to give land-reform farmers and the officers who support them earlier warning of infrastructure failure.

Why this matters

A note from the founder

"I grew up on a farm in the Eastern Cape. I saw how one broken pump, a leaking line or a fallen fence could undo months of work, and how often help arrived only after the damage was done. FarmPulse comes from that memory. Farmers should not have to find out about a failure by losing a crop or an animal."

Chantelle Palesa Malahlela, Founder and Director

Why we think it can work

Small early warnings are cheap compared with a lost season. A human officer stays in charge, which matches how support programmes already operate. Every warning, approval and repair is recorded, so funders can see what was done and why.

Farmers

Earlier notice of wear, leaks and structural change, so repairs are planned rather than emergencies.

Farm support officers

One prioritised board across many farms instead of driving out to find problems. Drafted work orders to approve or reject.

Communities

Farms that keep running protect local food supply, jobs and the value of land reform investment.

Funders and government

Evidence-backed funding requests and an audit trail for infrastructure spending.

These are design goals for the product, not measured results. Impact will be measured in pilots with consenting farms.

Contact

Write to the address that matches your need.

Send us a message

Quorvex Labs (Pty) Ltd · 34 Squirrel Street, Villa Liza, Boksburg, Gauteng, 1459

Privacy, terms and disclaimers

Privacy notice (POPIA)

Responsible party: Quorvex Labs (Pty) Ltd, reg. no. 2026/723698/07. This site has no accounts, forms or tracking cookies. If you email us we keep your message and details only to reply and for our records, and we do not sell personal information. The demo sign-in stores nothing. Fonts load from Google Fonts and weather from Open-Meteo, so those services may see your IP address. Request access, correction or deletion at info@quorvexlabs.co.za. You may also complain to the Information Regulator of South Africa.

Data ownership

Farmers own their farm data. It would be shared only with their consent. No real farm data is held by this prototype.

Disclaimers

All charts, sensor readings, predictions and AI outputs on this site are simulated and all farms in the demo are fictional. FarmPulse predictions are decision support, not guarantees, and do not replace physical inspection. Quorvex Labs is not affiliated with or endorsed by any government body or cloud provider named on this site. Third-party names belong to their owners.