Back to Work Project 01 / Ventures
Project Reforest AI
Year 2024 / present
Role Founder · Technical Lead
Stage In operation
Sector Climate Tech

Reforest AI.

Summary

Revenue protection infrastructure for forest-backed carbon assets. We help REDD+ project developers detect deforestation, reversals, and fraud risk before they erase years of credits.

Reforest AI
Fig.01 Reforest AI, Monitoring Concept
▶ / Reel Watch the demo.

The problem.

Carbon credits generated from protected forest are only worth what the forest is worth. When trees disappear, through illegal logging, political unrest, fire, or quiet boundary creep, the revenue disappears with them. Buyers quietly delist projects. Insurers raise premiums. Developers find out months after the damage is done, and by then the credits on sale have already lost credibility.

The industry needs an early-warning system that speaks the language of revenue, not science.

What we build.

Reforest AI ingests satellite imagery, on-ground sensor telemetry, and public registries, and turns them into a revenue-risk signal tailored to each REDD+ project. Developers see, in dollars, which hectares are at risk of reversal, and receive actionable alerts before credits go stale.

Detection Multi-source satellite + ground sensor fusion
Cadence Daily risk refresh
Output Hectare-scored revenue-at-risk feed
Integration REST API · Webhooks · Verra-aligned
Stack FastAPI · PostgreSQL · PyTorch · Supabase
Deployment AWS · Vercel · Docker

"We don't sell a dashboard. We sell the minute a developer knew their forest was at risk."

Approach.

The architecture is intentionally dull in the right places and opinionated where it matters. We separated the pipeline into three layers: a raw ingest tier that treats every data source as eventually-consistent, a scoring tier that converts change detection into financial risk, and a thin product surface that developers can integrate in a day.

Nothing exotic. The scoring layer is where the company lives, converting pixels into dollars for a specific project geometry, without overclaiming precision the data cannot support.

Fig.02 Pipeline / Ingest, Score, Surface

Fig.02 / Three-tier pipeline architecture: Ingest layer collects raw satellite and sensor data, Score layer computes hectare-weighted revenue-at-risk, Surface layer exposes alerts and APIs to developers.

Recognition.

Reforest AI was awarded Top Honors in the 2025 National Geographic Society Slingshot Challenge, a $10,000 grant, selected from 2,700+ entries across 96 countries. The project has been featured in Tech Cabal, TechPoint Africa, Disrupt Africa, and Hackster.io.

Award Nat Geo Slingshot, Top Honors
Grant $10,000
Field 2,700+ entries / 96 countries
Talk TEDx, The Future of Robotics

Status.

In operation. We are currently piloting with REDD+ developers and selectively onboarding partners. If you operate a carbon project and want to see your forest through Reforest AI's lens, get in touch.

Start a conversation →

Next Project 02

Air Clicks.