Taranis — leaf-level aerial imagery and TensorFlow ML across multiple continents

MENA (Israel origin), NA-US, EU, LATAM, SSA (multi-continent)

Content

Taranis is a precision agriculture intelligence platform using drone-captured leaf-level imagery plus satellite and plane imagery, processed through TensorFlow-based ML models trained on tens of millions of tagged images. Founded 2014 in Israel.

Concrete figures from Taranis / Google Cloud case study:

• More than 20 million acres managed using its intelligence platform.
• Active in Israel, US, Russia, Eastern Europe, South America (Argentina, Brazil).
• 30-person agronomy team tagging images to train ML models.
• Each drone flight collects ~10,000 images, 10–20MB each.

Deployment model. Drone-captured imagery uploaded to Google Cloud, ML models run on TensorFlow, insights delivered to farmers / agronomists. Cross-continental scale, industrial and large-farm operational scale.

Taranis’ framing of the problem. Quoted CTO Eli Bukchin: “The big problem farmers face is maintaining oversight on hundreds of thousands of acres: up to 40 percent of crops are routinely lost because of insects, crop disease, weeds, and nutrient deficiencies.” The platform positions itself as enabling earlier and more targeted intervention with fewer chemicals.

What this unit is doing in the taxonomy

Anchors the aerial robotics × computer vision × disease/pest/weed detection × on-farm open field cell at meaningful cross-continental scale. Distinct from Chinese drone manufacturers (DJI, XAG) — Taranis is a services-and-platform play on top of hardware, not a hardware vendor. Distinct from JD See & Spray — Taranis is detection and insight; See & Spray is actuation.

Why it matters for talks

• Strong example of the drone-captured imagery + cloud-ML + insight delivery deployment pattern that bridges Israeli-origin technology with multi-continental deployment.
• The 30-person agronomist team tagging images is a useful operational detail — the labour of training data is often invisible in vendor framing.
• Useful contrast with Chinese drone deployment (DJI hardware-led, government-incentivised) and with EU equipment-integration (CLAAS, AGCO PTx).

Critical context

• 20 million acres managed is a vendor-reported figure from the Google Cloud case study. Independent verification at scale not surfaced (G-020).
• The platform requires connectivity to upload large drone imagery files — Taranis’ own CTO names this as a challenge in Russia, Eastern Europe, South America. The unit shouldn’t be cited as deployed in low-connectivity contexts.
• The 40% routine crop loss figure is from Taranis’ own framing; UN FAO quotes vary. Worth flagging when used in talks.
• Taranis has been acquisitive — acquired Mavrx (US aerial imagery competitor) and other platforms. Useful context for industry consolidation but not central to the unit.